From 603d9882e054e39529b9876c52a94196bc6ee5ec Mon Sep 17 00:00:00 2001 From: jpan2 Date: Fri, 11 Mar 2022 09:54:40 -0800 Subject: [PATCH 1/2] Drop constraint for shap --- causalml/inference/tree/causaltree.html | 3588 ++++++ causalml/inference/tree/uplift.html | 13953 ++++++++++++++++++++++ requirements-tf.txt | 2 +- requirements.txt | 2 +- 4 files changed, 17543 insertions(+), 2 deletions(-) create mode 100644 causalml/inference/tree/causaltree.html create mode 100644 causalml/inference/tree/uplift.html diff --git a/causalml/inference/tree/causaltree.html b/causalml/inference/tree/causaltree.html new file mode 100644 index 00000000..a07ba4e7 --- /dev/null +++ b/causalml/inference/tree/causaltree.html @@ -0,0 +1,3588 @@ + + + + + + Cython: causaltree.pyx + + + +

Generated by Cython 0.29.24

+

+ Yellow lines hint at Python interaction.
+ Click on a line that starts with a "+" to see the C code that Cython generated for it. +

+

Raw output: causaltree.c

+
+001: # cython: cdivision=True
+
  __pyx_t_2 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_2) < 0) __PYX_ERR(0, 1, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 002: # cython: boundscheck=False
+
 003: # cython: wraparound=False
+
 004:
+
 005:
+
+006: import logging
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_logging, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 6, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_logging, __pyx_t_1) < 0) __PYX_ERR(0, 6, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+007: import numbers
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numbers, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 7, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_numbers, __pyx_t_1) < 0) __PYX_ERR(0, 7, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+008: import numpy as np
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 8, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 8, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+009: import pandas as pd
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_pandas, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 9, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_pd, __pyx_t_1) < 0) __PYX_ERR(0, 9, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 010:
+
+011: from math import ceil
+
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 11, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_ceil);
+  __Pyx_GIVEREF(__pyx_n_s_ceil);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_ceil);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_math, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 11, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_ceil); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 11, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_ceil, __pyx_t_1) < 0) __PYX_ERR(0, 11, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+012: from scipy.sparse import issparse
+
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 12, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s_issparse);
+  __Pyx_GIVEREF(__pyx_n_s_issparse);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_issparse);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_scipy_sparse, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 12, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_issparse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 12, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_issparse, __pyx_t_2) < 0) __PYX_ERR(0, 12, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+013: from scipy.stats import norm
+
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 13, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_norm);
+  __Pyx_GIVEREF(__pyx_n_s_norm);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_norm);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_scipy_stats, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 13, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_norm); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 13, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_norm, __pyx_t_1) < 0) __PYX_ERR(0, 13, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 014: from sklearn.tree._criterion cimport RegressionCriterion
+
 015: from sklearn.tree._criterion cimport SIZE_t, DOUBLE_t
+
+016: from sklearn.tree._splitter import BestSplitter
+
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 16, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s_BestSplitter);
+  __Pyx_GIVEREF(__pyx_n_s_BestSplitter);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_BestSplitter);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn_tree__splitter, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 16, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_BestSplitter); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 16, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_BestSplitter, __pyx_t_2) < 0) __PYX_ERR(0, 16, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+017: from sklearn.tree._tree import DepthFirstTreeBuilder, DOUBLE, DTYPE, Tree
+
  __pyx_t_1 = PyList_New(4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_DepthFirstTreeBuilder);
+  __Pyx_GIVEREF(__pyx_n_s_DepthFirstTreeBuilder);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_DepthFirstTreeBuilder);
+  __Pyx_INCREF(__pyx_n_s_DOUBLE);
+  __Pyx_GIVEREF(__pyx_n_s_DOUBLE);
+  PyList_SET_ITEM(__pyx_t_1, 1, __pyx_n_s_DOUBLE);
+  __Pyx_INCREF(__pyx_n_s_DTYPE);
+  __Pyx_GIVEREF(__pyx_n_s_DTYPE);
+  PyList_SET_ITEM(__pyx_t_1, 2, __pyx_n_s_DTYPE);
+  __Pyx_INCREF(__pyx_n_s_Tree);
+  __Pyx_GIVEREF(__pyx_n_s_Tree);
+  PyList_SET_ITEM(__pyx_t_1, 3, __pyx_n_s_Tree);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_tree__tree, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DepthFirstTreeBuilder); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DepthFirstTreeBuilder, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DOUBLE, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DTYPE, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_Tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Tree, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+018: from sklearn.utils import check_array, check_random_state
+
  __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s_check_array);
+  __Pyx_GIVEREF(__pyx_n_s_check_array);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_check_array);
+  __Pyx_INCREF(__pyx_n_s_check_random_state);
+  __Pyx_GIVEREF(__pyx_n_s_check_random_state);
+  PyList_SET_ITEM(__pyx_t_2, 1, __pyx_n_s_check_random_state);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn_utils, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_check_array); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_array, __pyx_t_2) < 0) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_random_state, __pyx_t_2) < 0) __PYX_ERR(0, 18, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 019:
+
+020: from causalml.inference.meta.utils import check_treatment_vector
+
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_check_treatment_vector);
+  __Pyx_GIVEREF(__pyx_n_s_check_treatment_vector);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_check_treatment_vector);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_causalml_inference_meta_utils, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_treatment_vector); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_treatment_vector, __pyx_t_1) < 0) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 021:
+
+022: logger = logging.getLogger('causalml')
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_logging); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_getLogger); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_tuple__25, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_logger, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+/* … */
+  __pyx_tuple__25 = PyTuple_Pack(1, __pyx_n_s_causalml); if (unlikely(!__pyx_tuple__25)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__25);
+  __Pyx_GIVEREF(__pyx_tuple__25);
+
 023:
+
 024:
+
+025: cdef class CausalMSE(RegressionCriterion):
+
struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE {
+  struct __pyx_obj_7sklearn_4tree_10_criterion_RegressionCriterion __pyx_base;
+};
+/* … */
+struct __pyx_vtabstruct_8causalml_9inference_4tree_10causaltree_CausalMSE {
+  struct __pyx_vtabstruct_7sklearn_4tree_10_criterion_RegressionCriterion __pyx_base;
+};
+static struct __pyx_vtabstruct_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_vtabptr_8causalml_9inference_4tree_10causaltree_CausalMSE;
+
+
 026:     """Causal Tree mean squared error impurity criterion.
+
 027:
+
 028:         CausalTreeMSE = right_effect + left_effect
+
 029:
+
 030:         where,
+
 031:
+
 032:         effect = alpha * tau^2 - (1 - alpha) * (1 + train_to_est_ratio) * (VAR_tr / p + VAR_cont / (1 - p))
+
 033:     """
+
 034:
+
+035:     cdef void node_value(self, double* dest) nogil:
+
static void __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_node_value(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self, double *__pyx_v_dest) {
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
+  double __pyx_v_node_ct;
+  double __pyx_v_node_tr;
+  double __pyx_v_node_ct_sum;
+  double __pyx_v_node_tr_sum;
+  double __pyx_v_one_over_eps;
+/* … */
+  /* function exit code */
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.node_value", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
+  __pyx_L0:;
+}
+
 036:         """Compute the node value of samples[start:end] into dest."""
+
 037:
+
+038:         cdef SIZE_t start = self.start
+
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.start;
+  __pyx_v_start = __pyx_t_1;
+
+039:         cdef SIZE_t end = self.end
+
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.end;
+  __pyx_v_end = __pyx_t_1;
+
 040:
+
 041:         cdef SIZE_t i
+
 042:         cdef SIZE_t p
+
 043:         cdef DOUBLE_t is_treated
+
 044:         cdef DOUBLE_t y_ik
+
 045:
+
+046:         cdef SIZE_t* samples = self.samples
+
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.samples;
+  __pyx_v_samples = __pyx_t_2;
+
+047:         cdef DOUBLE_t* sample_weight = self.sample_weight
+
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
+  __pyx_v_sample_weight = __pyx_t_3;
+
 048:
+
+049:         cdef double node_ct = 0.0
+
  __pyx_v_node_ct = 0.0;
+
+050:         cdef double node_tr = 0.0
+
  __pyx_v_node_tr = 0.0;
+
+051:         cdef double node_ct_sum = 0.0
+
  __pyx_v_node_ct_sum = 0.0;
+
+052:         cdef double node_tr_sum = 0.0
+
  __pyx_v_node_tr_sum = 0.0;
+
+053:         cdef double one_over_eps = 1e5
+
  __pyx_v_one_over_eps = 1e5;
+
 054:
+
+055:         for p in range(start, end):
+
  __pyx_t_1 = __pyx_v_end;
+  __pyx_t_4 = __pyx_t_1;
+  for (__pyx_t_5 = __pyx_v_start; __pyx_t_5 < __pyx_t_4; __pyx_t_5+=1) {
+    __pyx_v_p = __pyx_t_5;
+
+056:             i = samples[p]
+
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
+
 057:
+
+058:             if sample_weight != NULL:
+
    __pyx_t_6 = ((__pyx_v_sample_weight != NULL) != 0);
+    if (__pyx_t_6) {
+/* … */
+    }
+
 059:                 # the weights of 1 and 1 + eps are used for treatment and control respectively
+
+060:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
+
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
+
 061:
+
 062:             # assume that there is only one output (k = 0)
+
+063:             y_ik = self.y[i, 0]
+
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 63, __pyx_L1_error)}
+    __pyx_t_7 = __pyx_v_i;
+    __pyx_t_8 = 0;
+    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_7 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_8)) )));
+
 064:
+
+065:             node_tr += is_treated
+
    __pyx_v_node_tr = (__pyx_v_node_tr + __pyx_v_is_treated);
+
+066:             node_ct += 1. - is_treated
+
    __pyx_v_node_ct = (__pyx_v_node_ct + (1. - __pyx_v_is_treated));
+
+067:             node_tr_sum += y_ik * is_treated
+
    __pyx_v_node_tr_sum = (__pyx_v_node_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
+
+068:             node_ct_sum += y_ik * (1. - is_treated)
+
    __pyx_v_node_ct_sum = (__pyx_v_node_ct_sum + (__pyx_v_y_ik * (1. - __pyx_v_is_treated)));
+  }
+
 069:
+
 070:         # save the average of treatment effects within a node as a value for the node
+
+071:         dest[0] = node_tr_sum / node_tr - node_ct_sum / node_ct
+
  (__pyx_v_dest[0]) = ((__pyx_v_node_tr_sum / __pyx_v_node_tr) - (__pyx_v_node_ct_sum / __pyx_v_node_ct));
+
 072:
+
+073:     cdef double node_impurity(self) nogil:
+
static double __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_node_impurity(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self) {
+  CYTHON_UNUSED double *__pyx_v_sum_total;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
+  double __pyx_v_node_tr;
+  double __pyx_v_node_ct;
+  double __pyx_v_node_sum;
+  double __pyx_v_node_tr_sum;
+  double __pyx_v_node_sq_sum;
+  double __pyx_v_node_tr_sq_sum;
+  double __pyx_v_tr_var;
+  double __pyx_v_ct_var;
+  double __pyx_v_one_over_eps;
+  double __pyx_v_node_tau;
+  double __pyx_r;
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.node_impurity", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
+  __pyx_r = 0;
+  __pyx_L0:;
+  return __pyx_r;
+}
+
 074:         """Evaluate the impurity of the current node, i.e. the impurity of
+
 075:            samples[start:end]."""
+
 076:
+
+077:         cdef double* sum_total = self.sum_total
+
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.sum_total;
+  __pyx_v_sum_total = __pyx_t_1;
+
 078:         cdef double impurity
+
+079:         cdef SIZE_t start = self.start
+
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.start;
+  __pyx_v_start = __pyx_t_2;
+
+080:         cdef SIZE_t end = self.end
+
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.end;
+  __pyx_v_end = __pyx_t_2;
+
 081:
+
 082:         cdef SIZE_t i
+
 083:         cdef SIZE_t p
+
 084:         cdef DOUBLE_t is_treated
+
 085:         cdef DOUBLE_t y_ik
+
 086:
+
+087:         cdef SIZE_t* samples = self.samples
+
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.samples;
+  __pyx_v_samples = __pyx_t_3;
+
+088:         cdef DOUBLE_t* sample_weight = self.sample_weight
+
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
+  __pyx_v_sample_weight = __pyx_t_4;
+
 089:
+
+090:         cdef double node_tr = 0.0
+
  __pyx_v_node_tr = 0.0;
+
+091:         cdef double node_ct = 0.0
+
  __pyx_v_node_ct = 0.0;
+
+092:         cdef double node_sum = self.sum_total[0]
+
  __pyx_v_node_sum = (__pyx_v_self->__pyx_base.__pyx_base.sum_total[0]);
+
+093:         cdef double node_tr_sum = 0.0
+
  __pyx_v_node_tr_sum = 0.0;
+
+094:         cdef double node_sq_sum = 0.0
+
  __pyx_v_node_sq_sum = 0.0;
+
+095:         cdef double node_tr_sq_sum = 0.0
+
  __pyx_v_node_tr_sq_sum = 0.0;
+
 096:         cdef double tr_var
+
 097:         cdef double ct_var
+
+098:         cdef double one_over_eps = 1e5
+
  __pyx_v_one_over_eps = 1e5;
+
 099:
+
+100:         for p in range(start, end):
+
  __pyx_t_2 = __pyx_v_end;
+  __pyx_t_5 = __pyx_t_2;
+  for (__pyx_t_6 = __pyx_v_start; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
+    __pyx_v_p = __pyx_t_6;
+
+101:             i = samples[p]
+
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
+
 102:
+
+103:             if sample_weight != NULL:
+
    __pyx_t_7 = ((__pyx_v_sample_weight != NULL) != 0);
+    if (__pyx_t_7) {
+/* … */
+    }
+
 104:                 # the weights of 1 and 1 + eps are used for treatment and control respectively
+
+105:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
+
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
+
 106:
+
 107:             # assume that there is only one output (k = 0)
+
+108:             y_ik = self.y[i, 0]
+
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 108, __pyx_L1_error)}
+    __pyx_t_8 = __pyx_v_i;
+    __pyx_t_9 = 0;
+    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_8 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_9)) )));
+
 109:
+
+110:             node_tr += is_treated
+
    __pyx_v_node_tr = (__pyx_v_node_tr + __pyx_v_is_treated);
+
+111:             node_ct += (1. - is_treated)
+
    __pyx_v_node_ct = (__pyx_v_node_ct + (1. - __pyx_v_is_treated));
+
+112:             node_tr_sum += y_ik * is_treated
+
    __pyx_v_node_tr_sum = (__pyx_v_node_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
+
+113:             node_sq_sum += y_ik * y_ik
+
    __pyx_v_node_sq_sum = (__pyx_v_node_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
+
+114:             node_tr_sq_sum += y_ik * y_ik * is_treated
+
    __pyx_v_node_tr_sq_sum = (__pyx_v_node_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
+  }
+
 115:
+
+116:         node_tau = node_tr_sum / node_tr - (node_sum - node_tr_sum) / node_ct
+
  __pyx_v_node_tau = ((__pyx_v_node_tr_sum / __pyx_v_node_tr) - ((__pyx_v_node_sum - __pyx_v_node_tr_sum) / __pyx_v_node_ct));
+
+117:         tr_var = node_tr_sq_sum / node_tr - node_tr_sum * node_tr_sum / (node_tr * node_tr)
+
  __pyx_v_tr_var = ((__pyx_v_node_tr_sq_sum / __pyx_v_node_tr) - ((__pyx_v_node_tr_sum * __pyx_v_node_tr_sum) / (__pyx_v_node_tr * __pyx_v_node_tr)));
+
+118:         ct_var = ((node_sq_sum - node_tr_sq_sum) / node_ct -
+
  __pyx_v_ct_var = (((__pyx_v_node_sq_sum - __pyx_v_node_tr_sq_sum) / __pyx_v_node_ct) - (((__pyx_v_node_sum - __pyx_v_node_tr_sum) * (__pyx_v_node_sum - __pyx_v_node_tr_sum)) / (__pyx_v_node_ct * __pyx_v_node_ct)));
+
 119:                   (node_sum - node_tr_sum) * (node_sum - node_tr_sum) / (node_ct * node_ct))
+
 120:
+
+121:         return  (tr_var / node_tr + ct_var / node_ct) - node_tau * node_tau
+
  __pyx_r = (((__pyx_v_tr_var / __pyx_v_node_tr) + (__pyx_v_ct_var / __pyx_v_node_ct)) - (__pyx_v_node_tau * __pyx_v_node_tau));
+  goto __pyx_L0;
+
 122:
+
 123:
+
+124:     cdef void children_impurity(self, double* impurity_left, double* impurity_right) nogil:
+
static void __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_children_impurity(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self, double *__pyx_v_impurity_left, double *__pyx_v_impurity_right) {
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_pos;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
+  double *__pyx_v_sum_left;
+  double *__pyx_v_sum_right;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
+  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
+  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
+  double __pyx_v_right_tr;
+  double __pyx_v_right_ct;
+  double __pyx_v_right_sum;
+  double __pyx_v_right_tr_sum;
+  double __pyx_v_right_sq_sum;
+  double __pyx_v_right_tr_sq_sum;
+  double __pyx_v_right_tr_var;
+  double __pyx_v_right_ct_var;
+  double __pyx_v_left_tr;
+  double __pyx_v_left_ct;
+  double __pyx_v_left_sum;
+  double __pyx_v_left_tr_sum;
+  double __pyx_v_left_sq_sum;
+  double __pyx_v_left_tr_sq_sum;
+  double __pyx_v_left_tr_var;
+  double __pyx_v_left_ct_var;
+  double __pyx_v_one_over_eps;
+  double __pyx_v_right_tau;
+  double __pyx_v_left_tau;
+/* … */
+  /* function exit code */
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.children_impurity", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
+  __pyx_L0:;
+}
+
 125:         """Evaluate the impurity in children nodes, i.e. the impurity of the
+
 126:            left child (samples[start:pos]) and the impurity the right child
+
 127:            (samples[pos:end])."""
+
 128:
+
+129:         cdef DOUBLE_t* sample_weight = self.sample_weight
+
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
+  __pyx_v_sample_weight = __pyx_t_1;
+
+130:         cdef SIZE_t* samples = self.samples
+
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.samples;
+  __pyx_v_samples = __pyx_t_2;
+
+131:         cdef SIZE_t start = self.start
+
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.start;
+  __pyx_v_start = __pyx_t_3;
+
+132:         cdef SIZE_t pos = self.pos
+
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.pos;
+  __pyx_v_pos = __pyx_t_3;
+
+133:         cdef SIZE_t end = self.end
+
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.end;
+  __pyx_v_end = __pyx_t_3;
+
 134:
+
+135:         cdef double* sum_left = self.sum_left
+
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sum_left;
+  __pyx_v_sum_left = __pyx_t_4;
+
+136:         cdef double* sum_right = self.sum_right
+
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sum_right;
+  __pyx_v_sum_right = __pyx_t_4;
+
 137:
+
 138:         cdef SIZE_t i
+
 139:         cdef SIZE_t p
+
 140:         cdef DOUBLE_t is_treated
+
 141:         cdef DOUBLE_t y_ik
+
 142:
+
+143:         cdef double right_tr = 0.0
+
  __pyx_v_right_tr = 0.0;
+
+144:         cdef double right_ct = 0.0
+
  __pyx_v_right_ct = 0.0;
+
+145:         cdef double right_sum = 0.0
+
  __pyx_v_right_sum = 0.0;
+
+146:         cdef double right_tr_sum = 0.0
+
  __pyx_v_right_tr_sum = 0.0;
+
+147:         cdef double right_sq_sum = 0.0
+
  __pyx_v_right_sq_sum = 0.0;
+
+148:         cdef double right_tr_sq_sum = 0.0
+
  __pyx_v_right_tr_sq_sum = 0.0;
+
 149:         cdef double right_tr_var
+
 150:         cdef double right_ct_var
+
 151:
+
+152:         cdef double left_tr = 0.0
+
  __pyx_v_left_tr = 0.0;
+
+153:         cdef double left_ct = 0.0
+
  __pyx_v_left_ct = 0.0;
+
+154:         cdef double left_sum = 0.0
+
  __pyx_v_left_sum = 0.0;
+
+155:         cdef double left_tr_sum = 0.0
+
  __pyx_v_left_tr_sum = 0.0;
+
+156:         cdef double left_sq_sum = 0.0
+
  __pyx_v_left_sq_sum = 0.0;
+
+157:         cdef double left_tr_sq_sum = 0.0
+
  __pyx_v_left_tr_sq_sum = 0.0;
+
 158:         cdef double left_tr_var
+
 159:         cdef double left_ct_var
+
 160:
+
+161:         cdef double one_over_eps = 1e5
+
  __pyx_v_one_over_eps = 1e5;
+
 162:
+
+163:         for p in range(start, end):
+
  __pyx_t_3 = __pyx_v_end;
+  __pyx_t_5 = __pyx_t_3;
+  for (__pyx_t_6 = __pyx_v_start; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
+    __pyx_v_p = __pyx_t_6;
+
+164:             i = samples[p]
+
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
+
 165:
+
+166:             if sample_weight != NULL:
+
    __pyx_t_7 = ((__pyx_v_sample_weight != NULL) != 0);
+    if (__pyx_t_7) {
+/* … */
+    }
+
 167:                 # the weights of 1 and 1 + eps are used for control and treatment respectively
+
+168:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
+
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
+
 169:
+
 170:             # assume that there is only one output (k = 0)
+
+171:             y_ik = self.y[i, 0]
+
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 171, __pyx_L1_error)}
+    __pyx_t_8 = __pyx_v_i;
+    __pyx_t_9 = 0;
+    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_8 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_9)) )));
+
 172:
+
+173:             if p < pos:
+
    __pyx_t_7 = ((__pyx_v_p < __pyx_v_pos) != 0);
+    if (__pyx_t_7) {
+/* … */
+      goto __pyx_L6;
+    }
+
+174:                 left_tr += is_treated
+
      __pyx_v_left_tr = (__pyx_v_left_tr + __pyx_v_is_treated);
+
+175:                 left_ct += 1. - is_treated
+
      __pyx_v_left_ct = (__pyx_v_left_ct + (1. - __pyx_v_is_treated));
+
+176:                 left_sum += y_ik
+
      __pyx_v_left_sum = (__pyx_v_left_sum + __pyx_v_y_ik);
+
+177:                 left_tr_sum += y_ik * is_treated
+
      __pyx_v_left_tr_sum = (__pyx_v_left_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
+
+178:                 left_sq_sum += y_ik * y_ik
+
      __pyx_v_left_sq_sum = (__pyx_v_left_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
+
+179:                 left_tr_sq_sum += y_ik * y_ik * is_treated
+
      __pyx_v_left_tr_sq_sum = (__pyx_v_left_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
+
 180:             else:
+
+181:                 right_tr += is_treated
+
    /*else*/ {
+      __pyx_v_right_tr = (__pyx_v_right_tr + __pyx_v_is_treated);
+
+182:                 right_ct += 1. - is_treated
+
      __pyx_v_right_ct = (__pyx_v_right_ct + (1. - __pyx_v_is_treated));
+
+183:                 right_sum += y_ik
+
      __pyx_v_right_sum = (__pyx_v_right_sum + __pyx_v_y_ik);
+
+184:                 right_tr_sum += y_ik * is_treated
+
      __pyx_v_right_tr_sum = (__pyx_v_right_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
+
+185:                 right_sq_sum += y_ik * y_ik
+
      __pyx_v_right_sq_sum = (__pyx_v_right_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
+
+186:                 right_tr_sq_sum += y_ik * y_ik * is_treated
+
      __pyx_v_right_tr_sq_sum = (__pyx_v_right_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
+    }
+    __pyx_L6:;
+  }
+
 187:
+
+188:         right_tau = right_tr_sum / right_tr - (sum_right[0] - right_tr_sum) / right_ct
+
  __pyx_v_right_tau = ((__pyx_v_right_tr_sum / __pyx_v_right_tr) - (((__pyx_v_sum_right[0]) - __pyx_v_right_tr_sum) / __pyx_v_right_ct));
+
+189:         right_tr_var = right_tr_sq_sum / right_tr - right_tr_sum * right_tr_sum / (right_tr * right_tr)
+
  __pyx_v_right_tr_var = ((__pyx_v_right_tr_sq_sum / __pyx_v_right_tr) - ((__pyx_v_right_tr_sum * __pyx_v_right_tr_sum) / (__pyx_v_right_tr * __pyx_v_right_tr)));
+
+190:         right_ct_var = ((right_sq_sum - right_tr_sq_sum) / right_ct -
+
  __pyx_v_right_ct_var = (((__pyx_v_right_sq_sum - __pyx_v_right_tr_sq_sum) / __pyx_v_right_ct) - (((__pyx_v_right_sum - __pyx_v_right_tr_sum) * (__pyx_v_right_sum - __pyx_v_right_tr_sum)) / (__pyx_v_right_ct * __pyx_v_right_ct)));
+
 191:                          (right_sum - right_tr_sum) * (right_sum - right_tr_sum) / (right_ct * right_ct))
+
 192:
+
+193:         left_tau = left_tr_sum / left_tr - (sum_left[0] - left_tr_sum) / left_ct
+
  __pyx_v_left_tau = ((__pyx_v_left_tr_sum / __pyx_v_left_tr) - (((__pyx_v_sum_left[0]) - __pyx_v_left_tr_sum) / __pyx_v_left_ct));
+
+194:         left_tr_var = left_tr_sq_sum / left_tr - left_tr_sum * left_tr_sum / (left_tr * left_tr)
+
  __pyx_v_left_tr_var = ((__pyx_v_left_tr_sq_sum / __pyx_v_left_tr) - ((__pyx_v_left_tr_sum * __pyx_v_left_tr_sum) / (__pyx_v_left_tr * __pyx_v_left_tr)));
+
+195:         left_ct_var = ((left_sq_sum - left_tr_sq_sum) / left_ct -
+
  __pyx_v_left_ct_var = (((__pyx_v_left_sq_sum - __pyx_v_left_tr_sq_sum) / __pyx_v_left_ct) - (((__pyx_v_left_sum - __pyx_v_left_tr_sum) * (__pyx_v_left_sum - __pyx_v_left_tr_sum)) / (__pyx_v_left_ct * __pyx_v_left_ct)));
+
 196:                         (left_sum - left_tr_sum) * (left_sum - left_tr_sum) / (left_ct * left_ct))
+
 197:
+
+198:         impurity_left[0] = (left_tr_var / left_tr + left_ct_var / left_ct) - left_tau * left_tau
+
  (__pyx_v_impurity_left[0]) = (((__pyx_v_left_tr_var / __pyx_v_left_tr) + (__pyx_v_left_ct_var / __pyx_v_left_ct)) - (__pyx_v_left_tau * __pyx_v_left_tau));
+
+199:         impurity_right[0] = (right_tr_var / right_tr + right_ct_var / right_ct) - right_tau * right_tau
+
  (__pyx_v_impurity_right[0]) = (((__pyx_v_right_tr_var / __pyx_v_right_tr) + (__pyx_v_right_ct_var / __pyx_v_right_ct)) - (__pyx_v_right_tau * __pyx_v_right_tau));
+
 200:
+
 201:
+
+202: class CausalTreeRegressor:
+
  __pyx_t_2 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_CausalTreeRegressor, __pyx_n_s_CausalTreeRegressor, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_kp_s_A_Causal_Tree_regressor_class_Th); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 202, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+/* … */
+  __pyx_t_1 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_CausalTreeRegressor, __pyx_empty_tuple, __pyx_t_2, NULL, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 202, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_CausalTreeRegressor, __pyx_t_1) < 0) __PYX_ERR(0, 202, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 203:     """A Causal Tree regressor class.
+
 204:
+
 205:     The Causal Tree is a decision tree regressor with a split criteria for treatment effects instead of
+
 206:     outputs.
+
 207:
+
 208:     Details are available at Athey and Imbens (2015) (https://arxiv.org/abs/1504.01132)
+
 209:     """
+
+210:     def __init__(self, ate_alpha=.05, control_name=0, max_depth=None,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__[] = "Initialize a Causal Tree\n\n        Args:\n            ate_alpha (float, optional): the confidence level alpha of the ATE estimate\n            control_name (str or int, optional): name of control group\n            max_depth (int, optional): the maximum depth of tree\n            min_samples_leaf (int, optional): the minimum number of samples in leaves\n            random_state (int or np.RandomState, optional): a random seed or a random state\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_ate_alpha = 0;
+  PyObject *__pyx_v_control_name = 0;
+  PyObject *__pyx_v_max_depth = 0;
+  PyObject *__pyx_v_min_samples_leaf = 0;
+  PyObject *__pyx_v_random_state = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_ate_alpha,&__pyx_n_s_control_name,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_random_state,0};
+    PyObject* values[6] = {0,0,0,0,0,0};
+    values[1] = ((PyObject *)((PyObject*)__pyx_float__05));
+    values[2] = ((PyObject *)((PyObject *)__pyx_int_0));
+    values[3] = ((PyObject *)((PyObject *)Py_None));
+    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
+/* … */
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_ate_alpha, PyObject *__pyx_v_control_name, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_random_state) {
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__", 0);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__26 = PyTuple_Pack(6, __pyx_n_s_self, __pyx_n_s_ate_alpha, __pyx_n_s_control_name, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_random_state); if (unlikely(!__pyx_tuple__26)) __PYX_ERR(0, 210, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__26);
+  __Pyx_GIVEREF(__pyx_tuple__26);
+  __pyx_codeobj__27 = (PyObject*)__Pyx_PyCode_New(6, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__26, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_init, 210, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__27)) __PYX_ERR(0, 210, __pyx_L1_error)
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__, 0, __pyx_n_s_CausalTreeRegressor___init, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__27)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 210, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__28);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init, __pyx_t_1) < 0) __PYX_ERR(0, 210, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_tuple__28 = PyTuple_Pack(5, ((PyObject*)__pyx_float__05), ((PyObject *)__pyx_int_0), ((PyObject *)Py_None), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__28)) __PYX_ERR(0, 210, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__28);
+  __Pyx_GIVEREF(__pyx_tuple__28);
+
+211:                  min_samples_leaf=100, random_state=None):
+
    values[5] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_ate_alpha);
+          if (value) { values[1] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name);
+          if (value) { values[2] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
+          if (value) { values[3] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
+          if (value) { values[5] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 210, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_ate_alpha = values[1];
+    __pyx_v_control_name = values[2];
+    __pyx_v_max_depth = values[3];
+    __pyx_v_min_samples_leaf = values[4];
+    __pyx_v_random_state = values[5];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("__init__", 0, 1, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 210, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__(__pyx_self, __pyx_v_self, __pyx_v_ate_alpha, __pyx_v_control_name, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_random_state);
+
 212:         """Initialize a Causal Tree
+
 213:
+
 214:         Args:
+
 215:             ate_alpha (float, optional): the confidence level alpha of the ATE estimate
+
 216:             control_name (str or int, optional): name of control group
+
 217:             max_depth (int, optional): the maximum depth of tree
+
 218:             min_samples_leaf (int, optional): the minimum number of samples in leaves
+
 219:             random_state (int or np.RandomState, optional): a random seed or a random state
+
 220:         """
+
+221:         self.ate_alpha = ate_alpha
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha, __pyx_v_ate_alpha) < 0) __PYX_ERR(0, 221, __pyx_L1_error)
+
+222:         self.control_name = control_name
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 222, __pyx_L1_error)
+
+223:         self.max_depth = max_depth
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 223, __pyx_L1_error)
+
+224:         self.min_samples_leaf = min_samples_leaf
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 224, __pyx_L1_error)
+
+225:         self.random_state = random_state
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 225, __pyx_L1_error)
+
 226:
+
+227:         self._classes = {}
+
  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 227, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_1) < 0) __PYX_ERR(0, 227, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+228:         self.tree = None
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_tree, Py_None) < 0) __PYX_ERR(0, 228, __pyx_L1_error)
+
 229:
+
+230:         self.eps = 1e-5
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_eps, __pyx_float_1eneg_5) < 0) __PYX_ERR(0, 230, __pyx_L1_error)
+
 231:
+
+232:     def fit(self, X, treatment, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit[] = "Fit the Causal Tree model\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n\n        Returns:\n            self (CausalTree object)\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 232, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 232, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 232, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 232, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 232, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_is_treatment = NULL;
+  PyObject *__pyx_v_w = NULL;
+  PyObject *__pyx_v_t_groups = NULL;
+  PyObject *__pyx_v_random_state = NULL;
+  PyObject *__pyx_v_n_samples = NULL;
+  PyObject *__pyx_v_n_features = NULL;
+  PyObject *__pyx_v_n_outputs = NULL;
+  PyObject *__pyx_v_min_samples_leaf = NULL;
+  PyObject *__pyx_v_max_depth = NULL;
+  PyObject *__pyx_v_splitter = NULL;
+  PyObject *__pyx_v_builder = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit", 0);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_INCREF(__pyx_v_y);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_XDECREF(__pyx_t_12);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_is_treatment);
+  __Pyx_XDECREF(__pyx_v_w);
+  __Pyx_XDECREF(__pyx_v_t_groups);
+  __Pyx_XDECREF(__pyx_v_random_state);
+  __Pyx_XDECREF(__pyx_v_n_samples);
+  __Pyx_XDECREF(__pyx_v_n_features);
+  __Pyx_XDECREF(__pyx_v_n_outputs);
+  __Pyx_XDECREF(__pyx_v_min_samples_leaf);
+  __Pyx_XDECREF(__pyx_v_max_depth);
+  __Pyx_XDECREF(__pyx_v_splitter);
+  __Pyx_XDECREF(__pyx_v_builder);
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XDECREF(__pyx_v_y);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__29 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_is_treatment, __pyx_n_s_w, __pyx_n_s_t_groups, __pyx_n_s_random_state, __pyx_n_s_n_samples, __pyx_n_s_n_features, __pyx_n_s_n_outputs, __pyx_n_s_min_samples_leaf, __pyx_n_s_max_depth, __pyx_n_s_splitter, __pyx_n_s_builder); if (unlikely(!__pyx_tuple__29)) __PYX_ERR(0, 232, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__29);
+  __Pyx_GIVEREF(__pyx_tuple__29);
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit, 0, __pyx_n_s_CausalTreeRegressor_fit, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__30)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 232, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_fit, __pyx_t_1) < 0) __PYX_ERR(0, 232, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_codeobj__30 = (PyObject*)__Pyx_PyCode_New(4, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__29, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_fit, 232, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__30)) __PYX_ERR(0, 232, __pyx_L1_error)
+
 233:         """Fit the Causal Tree model
+
 234:
+
 235:         Args:
+
 236:             X (np.matrix): a feature matrix
+
 237:             treatment (np.array): a treatment vector
+
 238:             y (np.array): an outcome vector
+
 239:
+
 240:         Returns:
+
 241:             self (CausalTree object)
+
 242:         """
+
+243:         check_treatment_vector(treatment, self.control_name)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_treatment_vector); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 243, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 243, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = NULL;
+  __pyx_t_5 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_5 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment, __pyx_t_3};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment, __pyx_t_3};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  } else
+  #endif
+  {
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 243, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    if (__pyx_t_4) {
+      __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_4); __pyx_t_4 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_treatment);
+    __Pyx_GIVEREF(__pyx_v_treatment);
+    PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_5, __pyx_v_treatment);
+    __Pyx_GIVEREF(__pyx_t_3);
+    PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+244:         is_treatment = treatment != self.control_name
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 244, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_t_1, Py_NE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 244, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_is_treatment = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+245:         w = is_treatment.astype(int)
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_is_treatment, __pyx_n_s_astype); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 245, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_6 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_6)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_6);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_6, ((PyObject *)(&PyInt_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_1, ((PyObject *)(&PyInt_Type)));
+  __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 245, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_w = __pyx_t_2;
+  __pyx_t_2 = 0;
+
 246:
+
+247:         t_groups = np.unique(treatment[is_treatment])
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 247, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_unique); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 247, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_is_treatment); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 247, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_6);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_6, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 247, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_v_t_groups = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+248:         self._classes[t_groups[0]] = 0
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 248, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_t_groups, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 248, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_t_6, __pyx_int_0) < 0)) __PYX_ERR(0, 248, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+
 249:
+
 250:         # input checking replicated from BaseDecisionTree.fit()
+
+251:         random_state = check_random_state(self.random_state)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 251, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 251, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_6 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 251, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_random_state = __pyx_t_6;
+  __pyx_t_6 = 0;
+
+252:         X = check_array(X, dtype=DTYPE, accept_sparse="csc")
+
  __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_check_array); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_v_X);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
+  __pyx_t_1 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_3) < 0) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_accept_sparse, __pyx_n_s_csc) < 0) __PYX_ERR(0, 252, __pyx_L1_error)
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_3);
+  __pyx_t_3 = 0;
+
+253:         y = check_array(y, ensure_2d=False, dtype=None)
+
  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_check_array); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 253, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 253, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_v_y);
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 253, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_ensure_2d, Py_False) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_dtype, Py_None) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
+  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 253, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
+  __pyx_t_6 = 0;
+
+254:         if issparse(X):
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_issparse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_1)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_1);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_6 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_1, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
+  __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  if (__pyx_t_7) {
+/* … */
+  }
+
+255:             X.sort_indices()
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_sort_indices); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 255, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_1 = NULL;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+      __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_2);
+      if (likely(__pyx_t_1)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+        __Pyx_INCREF(__pyx_t_1);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_2, function);
+      }
+    }
+    __pyx_t_6 = (__pyx_t_1) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+    __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
+    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 255, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+
 256:
+
+257:             if X.indices.dtype != np.intc or X.indptr.dtype != np.intc:
+
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_indices); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_dtype); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_intc); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = PyObject_RichCompare(__pyx_t_2, __pyx_t_1, Py_NE); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    if (!__pyx_t_8) {
+    } else {
+      __pyx_t_7 = __pyx_t_8;
+      goto __pyx_L5_bool_binop_done;
+    }
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_indptr); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_dtype); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_intc); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = PyObject_RichCompare(__pyx_t_1, __pyx_t_2, Py_NE); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 257, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_7 = __pyx_t_8;
+    __pyx_L5_bool_binop_done:;
+    if (unlikely(__pyx_t_7)) {
+/* … */
+    }
+
+258:                 raise ValueError("No support for np.int64 index based "
+
      __pyx_t_6 = __Pyx_PyObject_Call(__pyx_builtin_ValueError, __pyx_tuple__3, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 258, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_6);
+      __Pyx_Raise(__pyx_t_6, 0, 0, 0);
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __PYX_ERR(0, 258, __pyx_L1_error)
+/* … */
+  __pyx_tuple__3 = PyTuple_Pack(1, __pyx_kp_s_No_support_for_np_int64_index_ba); if (unlikely(!__pyx_tuple__3)) __PYX_ERR(0, 258, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__3);
+  __Pyx_GIVEREF(__pyx_tuple__3);
+
 259:                                  "sparse matrices")
+
 260:
+
+261:         y = np.atleast_1d(y)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 261, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_atleast_1d); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 261, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_6 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_2, __pyx_v_y) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_y);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 261, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
+  __pyx_t_6 = 0;
+
+262:         if y.ndim == 1:
+
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_ndim); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 262, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_1 = __Pyx_PyInt_EqObjC(__pyx_t_6, __pyx_int_1, 1, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 262, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 262, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (__pyx_t_7) {
+/* … */
+  }
+
+263:             y = np.reshape(y, (-1, 1))
+
    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 263, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_reshape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 263, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = NULL;
+    __pyx_t_5 = 0;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
+      if (likely(__pyx_t_6)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+        __Pyx_INCREF(__pyx_t_6);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_2, function);
+        __pyx_t_5 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_2)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_y, __pyx_tuple__4};
+      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_1);
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_y, __pyx_tuple__4};
+      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_1);
+    } else
+    #endif
+    {
+      __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 263, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      if (__pyx_t_6) {
+        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
+      }
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_5, __pyx_v_y);
+      __Pyx_INCREF(__pyx_tuple__4);
+      __Pyx_GIVEREF(__pyx_tuple__4);
+      PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_tuple__4);
+      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_1);
+    __pyx_t_1 = 0;
+/* … */
+  __pyx_tuple__4 = PyTuple_Pack(2, __pyx_int_neg_1, __pyx_int_1); if (unlikely(!__pyx_tuple__4)) __PYX_ERR(0, 263, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__4);
+  __Pyx_GIVEREF(__pyx_tuple__4);
+
+264:         if getattr(y, "dtype", None) != DOUBLE or not y.flags.contiguous:
+
  __pyx_t_1 = __Pyx_GetAttr3(__pyx_v_y, __pyx_n_s_dtype, Py_None); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = PyObject_RichCompare(__pyx_t_1, __pyx_t_2, Py_NE); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (!__pyx_t_8) {
+  } else {
+    __pyx_t_7 = __pyx_t_8;
+    goto __pyx_L9_bool_binop_done;
+  }
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_flags); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_contiguous); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 264, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_9 = ((!__pyx_t_8) != 0);
+  __pyx_t_7 = __pyx_t_9;
+  __pyx_L9_bool_binop_done:;
+  if (__pyx_t_7) {
+/* … */
+  }
+
+265:             y = np.ascontiguousarray(y, dtype=DOUBLE)
+
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_ascontiguousarray); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_6) < 0) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 265, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
+    __pyx_t_6 = 0;
+
+266:         n_samples, n_features = X.shape
+
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 266, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
+    PyObject* sequence = __pyx_t_6;
+    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+    if (unlikely(size != 2)) {
+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+      __PYX_ERR(0, 266, __pyx_L1_error)
+    }
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    if (likely(PyTuple_CheckExact(sequence))) {
+      __pyx_t_1 = PyTuple_GET_ITEM(sequence, 0); 
+      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 1); 
+    } else {
+      __pyx_t_1 = PyList_GET_ITEM(sequence, 0); 
+      __pyx_t_2 = PyList_GET_ITEM(sequence, 1); 
+    }
+    __Pyx_INCREF(__pyx_t_1);
+    __Pyx_INCREF(__pyx_t_2);
+    #else
+    __pyx_t_1 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 266, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_2 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 266, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    #endif
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  } else {
+    Py_ssize_t index = -1;
+    __pyx_t_3 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 266, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_10 = Py_TYPE(__pyx_t_3)->tp_iternext;
+    index = 0; __pyx_t_1 = __pyx_t_10(__pyx_t_3); if (unlikely(!__pyx_t_1)) goto __pyx_L11_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_1);
+    index = 1; __pyx_t_2 = __pyx_t_10(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L11_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_3), 2) < 0) __PYX_ERR(0, 266, __pyx_L1_error)
+    __pyx_t_10 = NULL;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    goto __pyx_L12_unpacking_done;
+    __pyx_L11_unpacking_failed:;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_10 = NULL;
+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+    __PYX_ERR(0, 266, __pyx_L1_error)
+    __pyx_L12_unpacking_done:;
+  }
+  __pyx_v_n_samples = __pyx_t_1;
+  __pyx_t_1 = 0;
+  __pyx_v_n_features = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+267:         n_outputs = y.shape[1]
+
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_shape); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 267, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_6, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 267, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_v_n_outputs = __pyx_t_2;
+  __pyx_t_2 = 0;
+
 268:
+
+269:         if isinstance(self.min_samples_leaf, numbers.Integral):
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 269, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_numbers); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 269, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_Integral); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 269, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_t_7 = PyObject_IsInstance(__pyx_t_2, __pyx_t_1); if (unlikely(__pyx_t_7 == ((int)-1))) __PYX_ERR(0, 269, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_9 = (__pyx_t_7 != 0);
+  if (__pyx_t_9) {
+/* … */
+    goto __pyx_L13;
+  }
+
+270:             if not 1 <= self.min_samples_leaf:
+
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 270, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_2 = PyObject_RichCompare(__pyx_int_1, __pyx_t_1, Py_LE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 270, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 270, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_7 = ((!__pyx_t_9) != 0);
+    if (unlikely(__pyx_t_7)) {
+/* … */
+    }
+
+271:                 raise ValueError("min_samples_leaf must be at least 1 "
+
      __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_ValueError, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 271, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __Pyx_Raise(__pyx_t_2, 0, 0, 0);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __PYX_ERR(0, 271, __pyx_L1_error)
+
 272:                                  "or in (0, 0.5], got %s"
+
+273:                                  % self.min_samples_leaf)
+
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 273, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_1 = __Pyx_PyString_FormatSafe(__pyx_kp_s_min_samples_leaf_must_be_at_leas, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 273, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+274:             min_samples_leaf = self.min_samples_leaf
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 274, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_v_min_samples_leaf = __pyx_t_2;
+    __pyx_t_2 = 0;
+
 275:         else:  # float
+
+276:             if not 0. < self.min_samples_leaf <= 0.5:
+
  /*else*/ {
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 276, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_1 = PyObject_RichCompare(__pyx_float_0_, __pyx_t_2, Py_LT); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 276, __pyx_L1_error)
+    if (__Pyx_PyObject_IsTrue(__pyx_t_1)) {
+      __Pyx_DECREF(__pyx_t_1);
+      __pyx_t_1 = PyObject_RichCompare(__pyx_t_2, __pyx_float_0_5, Py_LE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 276, __pyx_L1_error)
+    }
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 276, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_9 = ((!__pyx_t_7) != 0);
+    if (unlikely(__pyx_t_9)) {
+/* … */
+    }
+
+277:                 raise ValueError("min_samples_leaf must be at least 1 "
+
      __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_builtin_ValueError, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 277, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_Raise(__pyx_t_1, 0, 0, 0);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __PYX_ERR(0, 277, __pyx_L1_error)
+
 278:                                  "or in (0, 0.5], got %s"
+
+279:                                  % self.min_samples_leaf)
+
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 279, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_2 = __Pyx_PyString_FormatSafe(__pyx_kp_s_min_samples_leaf_must_be_at_leas, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 279, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+280:             min_samples_leaf = int(ceil(self.min_samples_leaf * n_samples))
+
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_ceil); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 280, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 280, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_3 = PyNumber_Multiply(__pyx_t_6, __pyx_v_n_samples); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 280, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
+      if (likely(__pyx_t_6)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+        __Pyx_INCREF(__pyx_t_6);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_2, function);
+      }
+    }
+    __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
+    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 280, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyNumber_Int(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 280, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_v_min_samples_leaf = __pyx_t_2;
+    __pyx_t_2 = 0;
+  }
+  __pyx_L13:;
+
+281:         max_depth = (np.iinfo(np.int32).max if self.max_depth is None
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_9 = (__pyx_t_1 == Py_None);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if ((__pyx_t_9 != 0)) {
+    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_iinfo); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_int32); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
+      __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_6);
+      if (likely(__pyx_t_3)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+        __Pyx_INCREF(__pyx_t_3);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_6, function);
+      }
+    }
+    __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_3, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_4);
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_max); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_2 = __pyx_t_6;
+    __pyx_t_6 = 0;
+  } else {
+
+282:                      else self.max_depth)
+
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 282, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_2 = __pyx_t_6;
+    __pyx_t_6 = 0;
+  }
+  __pyx_v_max_depth = __pyx_t_2;
+  __pyx_t_2 = 0;
+
 283:
+
+284:         self.tree = Tree(
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_Tree); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 284, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+/* … */
+  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 284, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_tree, __pyx_t_12) < 0) __PYX_ERR(0, 284, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+
+285:             n_features = n_features,
+
  __pyx_t_6 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 285, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_features, __pyx_v_n_features) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
+
 286:             # line below is taken from DecisionTreeRegressor.fit method source
+
 287:             #   which comments that the tree shouldn't need the n_classes parameter
+
 288:             #   but it apparently does
+
+289:             n_classes = np.array([1] * n_outputs, dtype=np.intp),
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_array); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_int_1);
+  __Pyx_GIVEREF(__pyx_int_1);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_int_1);
+  { PyObject* __pyx_temp = PyNumber_InPlaceMultiply(__pyx_t_1, __pyx_v_n_outputs); if (unlikely(!__pyx_temp)) __PYX_ERR(0, 289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_temp);
+    __Pyx_DECREF(__pyx_t_1);
+    __pyx_t_1 = __pyx_temp;
+  }
+  __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_GIVEREF(__pyx_t_1);
+  PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_1);
+  __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GetModuleGlobalName(__pyx_t_11, __pyx_n_s_np); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_intp); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_12) < 0) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 289, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_classes, __pyx_t_12) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+
+290:             n_outputs = n_outputs)
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_outputs, __pyx_v_n_outputs) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
+
+291:         splitter = BestSplitter(criterion = CausalMSE(1, X.shape[0]),
+
  __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_BestSplitter); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __pyx_t_6 = __Pyx_PyDict_NewPresized(5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_int_1);
+  __Pyx_GIVEREF(__pyx_int_1);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_int_1);
+  __Pyx_GIVEREF(__pyx_t_1);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_1);
+  __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_Call(((PyObject *)__pyx_ptype_8causalml_9inference_4tree_10causaltree_CausalMSE), __pyx_t_2, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_criterion, __pyx_t_1) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+/* … */
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_v_splitter = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+292:             max_features = n_features,
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_max_features, __pyx_v_n_features) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
+
+293:             min_samples_leaf = min_samples_leaf,
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_weight_leaf, __pyx_int_0) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
+
 294:             min_weight_leaf = 0, # from DecisionTreeRegressor default
+
+295:             random_state = random_state)
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
+
 296:         # hardcoded values below come from defaults values in
+
 297:         #   sklearn.tree._classes.DecisionTreeRegressor
+
+298:         builder = DepthFirstTreeBuilder(
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_DepthFirstTreeBuilder); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 298, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 298, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __pyx_v_builder = __pyx_t_12;
+  __pyx_t_12 = 0;
+
+299:             splitter = splitter,
+
  __pyx_t_6 = __Pyx_PyDict_NewPresized(7); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 299, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_splitter, __pyx_v_splitter) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_split, __pyx_int_2) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+
 300:             min_samples_split = 2,
+
+301:             min_samples_leaf = min_samples_leaf,
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_weight_leaf, __pyx_int_0) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+
 302:             min_weight_leaf = 0,
+
+303:             max_depth = max_depth,
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_impurity_decrease, __pyx_int_0) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+
 304:             min_impurity_decrease = 0,
+
+305:             min_impurity_split = float("-inf"))
+
  __pyx_t_12 = __Pyx_PyNumber_Float(__pyx_kp_s_inf); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 305, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_impurity_split, __pyx_t_12) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+
+306:         builder.build(
+
  __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_v_builder, __pyx_n_s_build); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+/* … */
+  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GIVEREF(__pyx_t_6);
+  PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_6);
+  __pyx_t_6 = 0;
+/* … */
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_t_1, __pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+307:             self.tree,
+
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 307, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+
+308:             X = X,
+
  __pyx_t_6 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_X, __pyx_v_X) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
+
+309:             y = y,
+
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_y, __pyx_v_y) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
+
+310:             sample_weight = 1 + self.eps * w)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_eps); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 310, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = PyNumber_Multiply(__pyx_t_2, __pyx_v_w); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 310, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyInt_AddCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 310, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_sample_weight, __pyx_t_2) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 311:
+
+312:         return self
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_self);
+  __pyx_r = __pyx_v_self;
+  goto __pyx_L0;
+
 313:
+
+314:     def predict(self, X):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict[] = "Predict treatment effects.\n\n        Args:\n            X (np.matrix): a feature matrix\n\n        Returns:\n            (numpy.ndarray): Predictions of treatment effects.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,0};
+    PyObject* values[2] = {0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, 1); __PYX_ERR(0, 314, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 314, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 314, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict(__pyx_self, __pyx_v_self, __pyx_v_X);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X) {
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict", 0);
+  __Pyx_INCREF(__pyx_v_X);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__31 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_X); if (unlikely(!__pyx_tuple__31)) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__31);
+  __Pyx_GIVEREF(__pyx_tuple__31);
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict, 0, __pyx_n_s_CausalTreeRegressor_predict, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__32)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_predict, __pyx_t_1) < 0) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_codeobj__32 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__31, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_predict, 314, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__32)) __PYX_ERR(0, 314, __pyx_L1_error)
+
 315:         """Predict treatment effects.
+
 316:
+
 317:         Args:
+
 318:             X (np.matrix): a feature matrix
+
 319:
+
 320:         Returns:
+
 321:             (numpy.ndarray): Predictions of treatment effects.
+
 322:         """
+
+323:         X = check_array(X, dtype=DTYPE, accept_sparse="csr")
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_check_array); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_v_X);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
+  __pyx_t_3 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_dtype, __pyx_t_4) < 0) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_accept_sparse, __pyx_n_s_csr) < 0) __PYX_ERR(0, 323, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_4);
+  __pyx_t_4 = 0;
+
+324:         return self.tree.predict(X).reshape((-1, 1))
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_predict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_3 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_2, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_X);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_reshape); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_tuple__4) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_tuple__4);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_r = __pyx_t_4;
+  __pyx_t_4 = 0;
+  goto __pyx_L0;
+
 325:
+
+326:     def fit_predict(self, X, treatment, y, return_ci=False, n_bootstraps=1000, bootstrap_size=10000, verbose=False):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict[] = "Fit the Causal Tree model and predict treatment effects.\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n            return_ci (bool): whether to return confidence intervals\n            n_bootstraps (int): number of bootstrap iterations\n            bootstrap_size (int): number of samples per bootstrap\n            verbose (str): whether to output progress logs\n\n        Returns:\n           (tuple):\n\n             - te (numpy.ndarray): Predictions of treatment effects.\n             - te_lower (numpy.ndarray, optional): lower bounds of treatment effects\n             - te_upper (numpy.ndarray, optional): upper bounds of treatment effects\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict = {"fit_predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_return_ci = 0;
+  PyObject *__pyx_v_n_bootstraps = 0;
+  PyObject *__pyx_v_bootstrap_size = 0;
+  PyObject *__pyx_v_verbose = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit_predict (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_return_ci,&__pyx_n_s_n_bootstraps,&__pyx_n_s_bootstrap_size,&__pyx_n_s_verbose,0};
+    PyObject* values[8] = {0,0,0,0,0,0,0,0};
+    values[4] = ((PyObject *)((PyObject *)Py_False));
+    values[5] = ((PyObject *)((PyObject *)__pyx_int_1000));
+    values[6] = ((PyObject *)((PyObject *)__pyx_int_10000));
+    values[7] = ((PyObject *)((PyObject *)Py_False));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 1); __PYX_ERR(0, 326, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 2); __PYX_ERR(0, 326, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 3); __PYX_ERR(0, 326, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_return_ci);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_bootstraps);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_bootstrap_size);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_verbose);
+          if (value) { values[7] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit_predict") < 0)) __PYX_ERR(0, 326, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_return_ci = values[4];
+    __pyx_v_n_bootstraps = values[5];
+    __pyx_v_bootstrap_size = values[6];
+    __pyx_v_verbose = values[7];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 326, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit_predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_return_ci, __pyx_v_n_bootstraps, __pyx_v_bootstrap_size, __pyx_v_verbose);
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_return_ci, PyObject *__pyx_v_n_bootstraps, PyObject *__pyx_v_bootstrap_size, PyObject *__pyx_v_verbose) {
+  PyObject *__pyx_v_te = NULL;
+  PyObject *__pyx_v_start = NULL;
+  PyObject *__pyx_v_te_bootstraps = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_te_b = NULL;
+  PyObject *__pyx_v_now = NULL;
+  PyObject *__pyx_v_lapsed = NULL;
+  PyObject *__pyx_v_te_lower = NULL;
+  PyObject *__pyx_v_te_upper = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit_predict", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_XDECREF(__pyx_t_12);
+  __Pyx_XDECREF(__pyx_t_13);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit_predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_te);
+  __Pyx_XDECREF(__pyx_v_start);
+  __Pyx_XDECREF(__pyx_v_te_bootstraps);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_te_b);
+  __Pyx_XDECREF(__pyx_v_now);
+  __Pyx_XDECREF(__pyx_v_lapsed);
+  __Pyx_XDECREF(__pyx_v_te_lower);
+  __Pyx_XDECREF(__pyx_v_te_upper);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__33 = PyTuple_Pack(17, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_return_ci, __pyx_n_s_n_bootstraps, __pyx_n_s_bootstrap_size, __pyx_n_s_verbose, __pyx_n_s_te, __pyx_n_s_start, __pyx_n_s_te_bootstraps, __pyx_n_s_i, __pyx_n_s_te_b, __pyx_n_s_now, __pyx_n_s_lapsed, __pyx_n_s_te_lower, __pyx_n_s_te_upper); if (unlikely(!__pyx_tuple__33)) __PYX_ERR(0, 326, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__33);
+  __Pyx_GIVEREF(__pyx_tuple__33);
+  __pyx_codeobj__34 = (PyObject*)__Pyx_PyCode_New(8, 0, 17, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__33, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_fit_predict, 326, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__34)) __PYX_ERR(0, 326, __pyx_L1_error)
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict, 0, __pyx_n_s_CausalTreeRegressor_fit_predict, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__34)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 326, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__35);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_fit_predict, __pyx_t_1) < 0) __PYX_ERR(0, 326, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_tuple__35 = PyTuple_Pack(4, ((PyObject *)Py_False), ((PyObject *)__pyx_int_1000), ((PyObject *)__pyx_int_10000), ((PyObject *)Py_False)); if (unlikely(!__pyx_tuple__35)) __PYX_ERR(0, 326, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__35);
+  __Pyx_GIVEREF(__pyx_tuple__35);
+
 327:         """Fit the Causal Tree model and predict treatment effects.
+
 328:
+
 329:         Args:
+
 330:             X (np.matrix): a feature matrix
+
 331:             treatment (np.array): a treatment vector
+
 332:             y (np.array): an outcome vector
+
 333:             return_ci (bool): whether to return confidence intervals
+
 334:             n_bootstraps (int): number of bootstrap iterations
+
 335:             bootstrap_size (int): number of samples per bootstrap
+
 336:             verbose (str): whether to output progress logs
+
 337:
+
 338:         Returns:
+
 339:            (tuple):
+
 340:
+
 341:              - te (numpy.ndarray): Predictions of treatment effects.
+
 342:              - te_lower (numpy.ndarray, optional): lower bounds of treatment effects
+
 343:              - te_upper (numpy.ndarray, optional): upper bounds of treatment effects
+
 344:         """
+
+345:         self.fit(X, treatment, y)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 345, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 345, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_X);
+    __Pyx_GIVEREF(__pyx_v_X);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
+    __Pyx_INCREF(__pyx_v_treatment);
+    __Pyx_GIVEREF(__pyx_v_treatment);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_treatment);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+346:         te = self.predict(X)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 346, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 346, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_te = __pyx_t_1;
+  __pyx_t_1 = 0;
+
 347:
+
+348:         if not return_ci:
+
  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_v_return_ci); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 348, __pyx_L1_error)
+  __pyx_t_7 = ((!__pyx_t_6) != 0);
+  if (__pyx_t_7) {
+/* … */
+  }
+
+349:             return te
+
    __Pyx_XDECREF(__pyx_r);
+    __Pyx_INCREF(__pyx_v_te);
+    __pyx_r = __pyx_v_te;
+    goto __pyx_L0;
+
 350:         else:
+
+351:             start = pd.datetime.today()
+
  /*else*/ {
+    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_pd); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 351, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_datetime); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 351, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_today); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 351, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = NULL;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+      if (likely(__pyx_t_5)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+        __Pyx_INCREF(__pyx_t_5);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_2, function);
+      }
+    }
+    __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 351, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_v_start = __pyx_t_1;
+    __pyx_t_1 = 0;
+
+352:             te_bootstraps = np.zeros(shape=(X.shape[0], n_bootstraps))
+
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_zeros); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_5, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_GIVEREF(__pyx_t_3);
+    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3);
+    __Pyx_INCREF(__pyx_v_n_bootstraps);
+    __Pyx_GIVEREF(__pyx_v_n_bootstraps);
+    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_n_bootstraps);
+    __pyx_t_3 = 0;
+    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_shape, __pyx_t_5) < 0) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_v_te_bootstraps = __pyx_t_5;
+    __pyx_t_5 = 0;
+
+353:             for i in range(n_bootstraps):
+
    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_v_n_bootstraps); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (likely(PyList_CheckExact(__pyx_t_5)) || PyTuple_CheckExact(__pyx_t_5)) {
+      __pyx_t_1 = __pyx_t_5; __Pyx_INCREF(__pyx_t_1); __pyx_t_8 = 0;
+      __pyx_t_9 = NULL;
+    } else {
+      __pyx_t_8 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 353, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_9 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 353, __pyx_L1_error)
+    }
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    for (;;) {
+      if (likely(!__pyx_t_9)) {
+        if (likely(PyList_CheckExact(__pyx_t_1))) {
+          if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_1)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_8); __Pyx_INCREF(__pyx_t_5); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 353, __pyx_L1_error)
+          #else
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_1, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_5);
+          #endif
+        } else {
+          if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_8); __Pyx_INCREF(__pyx_t_5); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 353, __pyx_L1_error)
+          #else
+          __pyx_t_5 = PySequence_ITEM(__pyx_t_1, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_5);
+          #endif
+        }
+      } else {
+        __pyx_t_5 = __pyx_t_9(__pyx_t_1);
+        if (unlikely(!__pyx_t_5)) {
+          PyObject* exc_type = PyErr_Occurred();
+          if (exc_type) {
+            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+            else __PYX_ERR(0, 353, __pyx_L1_error)
+          }
+          break;
+        }
+        __Pyx_GOTREF(__pyx_t_5);
+      }
+      __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_5);
+      __pyx_t_5 = 0;
+/* … */
+    }
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+354:                 te_b = self.bootstrap(X, treatment, y, size=bootstrap_size)
+
      __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_bootstrap); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 354, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 354, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_INCREF(__pyx_v_X);
+      __Pyx_GIVEREF(__pyx_v_X);
+      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
+      __Pyx_INCREF(__pyx_v_treatment);
+      __Pyx_GIVEREF(__pyx_v_treatment);
+      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_treatment);
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_y);
+      __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 354, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_size, __pyx_v_bootstrap_size) < 0) __PYX_ERR(0, 354, __pyx_L1_error)
+      __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 354, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_te_b, __pyx_t_10);
+      __pyx_t_10 = 0;
+
+355:                 te_bootstraps[:,i] = np.ravel(te_b)
+
      __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 355, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_ravel); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 355, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_3 = NULL;
+      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+        __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+        if (likely(__pyx_t_3)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+          __Pyx_INCREF(__pyx_t_3);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_2, function);
+        }
+      }
+      __pyx_t_10 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_te_b) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_te_b);
+      __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+      if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 355, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 355, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_INCREF(__pyx_slice__5);
+      __Pyx_GIVEREF(__pyx_slice__5);
+      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_slice__5);
+      __Pyx_INCREF(__pyx_v_i);
+      __Pyx_GIVEREF(__pyx_v_i);
+      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_i);
+      if (unlikely(PyObject_SetItem(__pyx_v_te_bootstraps, __pyx_t_2, __pyx_t_10) < 0)) __PYX_ERR(0, 355, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+/* … */
+  __pyx_slice__5 = PySlice_New(Py_None, Py_None, Py_None); if (unlikely(!__pyx_slice__5)) __PYX_ERR(0, 355, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_slice__5);
+  __Pyx_GIVEREF(__pyx_slice__5);
+
+356:                 if verbose:
+
      __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_verbose); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 356, __pyx_L1_error)
+      if (__pyx_t_7) {
+/* … */
+      }
+
+357:                     now = pd.datetime.today()
+
        __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_pd); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 357, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_datetime); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 357, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_today); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 357, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+          __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+          if (likely(__pyx_t_3)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+            __Pyx_INCREF(__pyx_t_3);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_2, function);
+          }
+        }
+        __pyx_t_10 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+        if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 357, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_now, __pyx_t_10);
+        __pyx_t_10 = 0;
+
+358:                     lapsed = (now-start).seconds / 60
+
        __pyx_t_10 = PyNumber_Subtract(__pyx_v_now, __pyx_v_start); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 358, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_10, __pyx_n_s_seconds); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 358, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+        __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_int_60); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 358, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_lapsed, __pyx_t_10);
+        __pyx_t_10 = 0;
+
+359:                     logger.info('{}/{} bootstraps completed. ({:.01f} min lapsed)'.format(i+1, n_bootstraps, lapsed))
+
        __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_logger); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_info); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 359, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_kp_s_bootstraps_completed_01f_min_la, __pyx_n_s_format); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 359, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __pyx_t_11 = __Pyx_PyInt_AddObjC(__pyx_v_i, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 359, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_12 = NULL;
+        __pyx_t_4 = 0;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
+          __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_5);
+          if (likely(__pyx_t_12)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+            __Pyx_INCREF(__pyx_t_12);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_5, function);
+            __pyx_t_4 = 1;
+          }
+        }
+        #if CYTHON_FAST_PYCALL
+        if (PyFunction_Check(__pyx_t_5)) {
+          PyObject *__pyx_temp[4] = {__pyx_t_12, __pyx_t_11, __pyx_v_n_bootstraps, __pyx_v_lapsed};
+          __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+          __Pyx_GOTREF(__pyx_t_2);
+          __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        } else
+        #endif
+        #if CYTHON_FAST_PYCCALL
+        if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
+          PyObject *__pyx_temp[4] = {__pyx_t_12, __pyx_t_11, __pyx_v_n_bootstraps, __pyx_v_lapsed};
+          __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+          __Pyx_GOTREF(__pyx_t_2);
+          __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        } else
+        #endif
+        {
+          __pyx_t_13 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 359, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_13);
+          if (__pyx_t_12) {
+            __Pyx_GIVEREF(__pyx_t_12); PyTuple_SET_ITEM(__pyx_t_13, 0, __pyx_t_12); __pyx_t_12 = NULL;
+          }
+          __Pyx_GIVEREF(__pyx_t_11);
+          PyTuple_SET_ITEM(__pyx_t_13, 0+__pyx_t_4, __pyx_t_11);
+          __Pyx_INCREF(__pyx_v_n_bootstraps);
+          __Pyx_GIVEREF(__pyx_v_n_bootstraps);
+          PyTuple_SET_ITEM(__pyx_t_13, 1+__pyx_t_4, __pyx_v_n_bootstraps);
+          __Pyx_INCREF(__pyx_v_lapsed);
+          __Pyx_GIVEREF(__pyx_v_lapsed);
+          PyTuple_SET_ITEM(__pyx_t_13, 2+__pyx_t_4, __pyx_v_lapsed);
+          __pyx_t_11 = 0;
+          __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_13, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_2);
+          __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
+        }
+        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        __pyx_t_5 = NULL;
+        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+          __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_3);
+          if (likely(__pyx_t_5)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+            __Pyx_INCREF(__pyx_t_5);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_3, function);
+          }
+        }
+        __pyx_t_10 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_5, __pyx_t_2) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_2);
+        __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 359, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
 360:
+
+361:             te_lower = np.percentile(te_bootstraps, (self.ate_alpha/2)*100, axis=1)
+
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_percentile); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = PyNumber_Multiply(__pyx_t_3, __pyx_int_100); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_INCREF(__pyx_v_te_bootstraps);
+    __Pyx_GIVEREF(__pyx_v_te_bootstraps);
+    PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_te_bootstraps);
+    __Pyx_GIVEREF(__pyx_t_1);
+    PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_t_1);
+    __pyx_t_1 = 0;
+    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 361, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 361, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_v_te_lower = __pyx_t_2;
+    __pyx_t_2 = 0;
+
+362:             te_upper = np.percentile(te_bootstraps, (1 - self.ate_alpha/2)*100, axis=1)
+
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_percentile); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = PyNumber_Multiply(__pyx_t_2, __pyx_int_100); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_v_te_bootstraps);
+    __Pyx_GIVEREF(__pyx_v_te_bootstraps);
+    PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_te_bootstraps);
+    __Pyx_GIVEREF(__pyx_t_3);
+    PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 362, __pyx_L1_error)
+    __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 362, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_v_te_upper = __pyx_t_10;
+    __pyx_t_10 = 0;
+
 363:
+
+364:             return (te, te_lower, te_upper)
+
    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_10 = PyTuple_New(3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 364, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __Pyx_INCREF(__pyx_v_te);
+    __Pyx_GIVEREF(__pyx_v_te);
+    PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_v_te);
+    __Pyx_INCREF(__pyx_v_te_lower);
+    __Pyx_GIVEREF(__pyx_v_te_lower);
+    PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_v_te_lower);
+    __Pyx_INCREF(__pyx_v_te_upper);
+    __Pyx_GIVEREF(__pyx_v_te_upper);
+    PyTuple_SET_ITEM(__pyx_t_10, 2, __pyx_v_te_upper);
+    __pyx_r = __pyx_t_10;
+    __pyx_t_10 = 0;
+    goto __pyx_L0;
+  }
+
 365:
+
+366:     def bootstrap(self, X, treatment, y, size=10000):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap[] = "Runs a single bootstrap.\n\n        Fits on bootstrapped sample, then predicts on whole population.\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n            size (int, optional): bootstrap sample size\n\n        Returns:\n            (np.array): bootstrap predictions\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap = {"bootstrap", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_size = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("bootstrap (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_size,0};
+    PyObject* values[5] = {0,0,0,0,0};
+    values[4] = ((PyObject *)((PyObject *)__pyx_int_10000));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 1); __PYX_ERR(0, 366, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 2); __PYX_ERR(0, 366, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 3); __PYX_ERR(0, 366, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_size);
+          if (value) { values[4] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "bootstrap") < 0)) __PYX_ERR(0, 366, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_size = values[4];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 366, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_size);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_size) {
+  PyObject *__pyx_v_idxs = NULL;
+  PyObject *__pyx_v_X_b = NULL;
+  PyObject *__pyx_v_treatment_b = NULL;
+  PyObject *__pyx_v_y_b = NULL;
+  PyObject *__pyx_v_te_b = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("bootstrap", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_idxs);
+  __Pyx_XDECREF(__pyx_v_X_b);
+  __Pyx_XDECREF(__pyx_v_treatment_b);
+  __Pyx_XDECREF(__pyx_v_y_b);
+  __Pyx_XDECREF(__pyx_v_te_b);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__36 = PyTuple_Pack(10, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_size, __pyx_n_s_idxs, __pyx_n_s_X_b, __pyx_n_s_treatment_b, __pyx_n_s_y_b, __pyx_n_s_te_b); if (unlikely(!__pyx_tuple__36)) __PYX_ERR(0, 366, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__36);
+  __Pyx_GIVEREF(__pyx_tuple__36);
+  __pyx_codeobj__37 = (PyObject*)__Pyx_PyCode_New(5, 0, 10, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__36, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_bootstrap, 366, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__37)) __PYX_ERR(0, 366, __pyx_L1_error)
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap, 0, __pyx_n_s_CausalTreeRegressor_bootstrap, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__37)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 366, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__38);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_bootstrap, __pyx_t_1) < 0) __PYX_ERR(0, 366, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_tuple__38 = PyTuple_Pack(1, ((PyObject *)__pyx_int_10000)); if (unlikely(!__pyx_tuple__38)) __PYX_ERR(0, 366, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__38);
+  __Pyx_GIVEREF(__pyx_tuple__38);
+
 367:         """Runs a single bootstrap.
+
 368:
+
 369:         Fits on bootstrapped sample, then predicts on whole population.
+
 370:
+
 371:         Args:
+
 372:             X (np.matrix): a feature matrix
+
 373:             treatment (np.array): a treatment vector
+
 374:             y (np.array): an outcome vector
+
 375:             size (int, optional): bootstrap sample size
+
 376:
+
 377:         Returns:
+
 378:             (np.array): bootstrap predictions
+
 379:         """
+
+380:         idxs = np.random.choice(np.arange(0, X.shape[0]), size=size)
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_random); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_choice); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_arange); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_3, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  __pyx_t_6 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_4);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_4, function);
+      __pyx_t_6 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_4)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_int_0, __pyx_t_5};
+    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_4)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_int_0, __pyx_t_5};
+    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  } else
+  #endif
+  {
+    __pyx_t_7 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 380, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_int_0);
+    __Pyx_GIVEREF(__pyx_int_0);
+    PyTuple_SET_ITEM(__pyx_t_7, 0+__pyx_t_6, __pyx_int_0);
+    __Pyx_GIVEREF(__pyx_t_5);
+    PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_6, __pyx_t_5);
+    __pyx_t_5 = 0;
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_GIVEREF(__pyx_t_2);
+  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
+  __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_size, __pyx_v_size) < 0) __PYX_ERR(0, 380, __pyx_L1_error)
+  __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 380, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_idxs = __pyx_t_7;
+  __pyx_t_7 = 0;
+
+381:         X_b = X[idxs]
+
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 381, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_v_X_b = __pyx_t_7;
+  __pyx_t_7 = 0;
+
+382:         treatment_b = treatment[idxs]
+
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 382, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_v_treatment_b = __pyx_t_7;
+  __pyx_t_7 = 0;
+
+383:         y_b = y[idxs]
+
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 383, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_v_y_b = __pyx_t_7;
+  __pyx_t_7 = 0;
+
+384:         self.fit(X=X_b, treatment=treatment_b, y=y_b)
+
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 384, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 384, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_X_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_treatment, __pyx_v_treatment_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_y, __pyx_v_y_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 384, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+385:         te_b = self.predict(X=X)
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_predict); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_X) < 0) __PYX_ERR(0, 385, __pyx_L1_error)
+  __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_te_b = __pyx_t_7;
+  __pyx_t_7 = 0;
+
+386:         return te_b
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_te_b);
+  __pyx_r = __pyx_v_te_b;
+  goto __pyx_L0;
+
 387:
+
+388:     def estimate_ate(self, X, treatment, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate[] = "Estimate the Average Treatment Effect (ATE).\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n\n        Returns:\n            The mean and confidence interval (LB, UB) of the ATE estimate.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate = {"estimate_ate", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("estimate_ate (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 1); __PYX_ERR(0, 388, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 2); __PYX_ERR(0, 388, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 3); __PYX_ERR(0, 388, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "estimate_ate") < 0)) __PYX_ERR(0, 388, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 388, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.estimate_ate", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_dhat = NULL;
+  PyObject *__pyx_v_te = NULL;
+  PyObject *__pyx_v_se = NULL;
+  PyObject *__pyx_v_te_lb = NULL;
+  PyObject *__pyx_v_te_ub = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("estimate_ate", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.estimate_ate", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_dhat);
+  __Pyx_XDECREF(__pyx_v_te);
+  __Pyx_XDECREF(__pyx_v_se);
+  __Pyx_XDECREF(__pyx_v_te_lb);
+  __Pyx_XDECREF(__pyx_v_te_ub);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__39 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_dhat, __pyx_n_s_te, __pyx_n_s_se, __pyx_n_s_te_lb, __pyx_n_s_te_ub); if (unlikely(!__pyx_tuple__39)) __PYX_ERR(0, 388, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__39);
+  __Pyx_GIVEREF(__pyx_tuple__39);
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate, 0, __pyx_n_s_CausalTreeRegressor_estimate_ate, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__40)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 388, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_estimate_ate, __pyx_t_1) < 0) __PYX_ERR(0, 388, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_codeobj__40 = (PyObject*)__Pyx_PyCode_New(4, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__39, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_estimate_ate, 388, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__40)) __PYX_ERR(0, 388, __pyx_L1_error)
+
 389:         """Estimate the Average Treatment Effect (ATE).
+
 390:
+
 391:         Args:
+
 392:             X (np.matrix): a feature matrix
+
 393:             treatment (np.array): a treatment vector
+
 394:             y (np.array): an outcome vector
+
 395:
+
 396:         Returns:
+
 397:             The mean and confidence interval (LB, UB) of the ATE estimate.
+
 398:         """
+
+399:         dhat = self.fit_predict(X, treatment, y)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 399, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 399, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_X);
+    __Pyx_GIVEREF(__pyx_v_X);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
+    __Pyx_INCREF(__pyx_v_treatment);
+    __Pyx_GIVEREF(__pyx_v_treatment);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_treatment);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_dhat = __pyx_t_1;
+  __pyx_t_1 = 0;
+
 400:
+
+401:         te = dhat.mean()
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_dhat, __pyx_n_s_mean); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 401, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 401, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_te = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+402:         se = dhat.std() / X.shape[0]
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_dhat, __pyx_n_s_std); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_se = __pyx_t_2;
+  __pyx_t_2 = 0;
+
 403:
+
+404:         te_lb = te - se * norm.ppf(1 - self.ate_alpha / 2)
+
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_norm); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_ppf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = PyNumber_Multiply(__pyx_v_se, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyNumber_Subtract(__pyx_v_te, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 404, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_te_lb = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+405:         te_ub = te + se * norm.ppf(1 - self.ate_alpha / 2)
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_norm); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_ppf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_5);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_5, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = PyNumber_Multiply(__pyx_v_se, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyNumber_Add(__pyx_v_te, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 405, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_te_ub = __pyx_t_2;
+  __pyx_t_2 = 0;
+
 406:
+
+407:         return te, te_lb, te_ub
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 407, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_te);
+  __Pyx_GIVEREF(__pyx_v_te);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_te);
+  __Pyx_INCREF(__pyx_v_te_lb);
+  __Pyx_GIVEREF(__pyx_v_te_lb);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_te_lb);
+  __Pyx_INCREF(__pyx_v_te_ub);
+  __Pyx_GIVEREF(__pyx_v_te_ub);
+  PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_te_ub);
+  __pyx_r = __pyx_t_2;
+  __pyx_t_2 = 0;
+  goto __pyx_L0;
+
diff --git a/causalml/inference/tree/uplift.html b/causalml/inference/tree/uplift.html new file mode 100644 index 00000000..eb93d071 --- /dev/null +++ b/causalml/inference/tree/uplift.html @@ -0,0 +1,13953 @@ + + + + + + Cython: uplift.pyx + + + +

Generated by Cython 0.29.24

+

+ Yellow lines hint at Python interaction.
+ Click on a line that starts with a "+" to see the C code that Cython generated for it. +

+

Raw output: uplift.c

+
+0001: # cython: cdivision=True
+
  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_1) < 0) __PYX_ERR(0, 1, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0002: # cython: boundscheck=False
+
 0003: # cython: wraparound=False
+
 0004: """
+
 0005: Forest of trees-based ensemble methods for Uplift modeling on Classification
+
 0006: Problem. Those methods include random forests and extremely randomized trees.
+
 0007:
+
 0008: The module structure is the following:
+
 0009: - The ``UpliftRandomForestClassifier`` base class implements different
+
 0010:   variants of uplift models based on random forest, with 'fit' and 'predict'
+
 0011:   method.
+
 0012: - The ``UpliftTreeClassifier`` base class implements the uplift trees (without
+
 0013:   Bootstrapping for random forest), this class is called within
+
 0014:   ``UpliftRandomForestClassifier`` for constructing random forest.
+
 0015: """
+
 0016:
+
 0017: # Authors: Zhenyu Zhao <zhenyuz@uber.com>
+
 0018: #          Totte Harinen <totte@uber.com>
+
 0019:
+
+0020: from collections import defaultdict
+
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_defaultdict);
+  __Pyx_GIVEREF(__pyx_n_s_defaultdict);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_defaultdict);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_collections, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_defaultdict, __pyx_t_1) < 0) __PYX_ERR(0, 20, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0021: import cython
+
+0022: from joblib import Parallel, delayed
+
  __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s_Parallel);
+  __Pyx_GIVEREF(__pyx_n_s_Parallel);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_Parallel);
+  __Pyx_INCREF(__pyx_n_s_delayed);
+  __Pyx_GIVEREF(__pyx_n_s_delayed);
+  PyList_SET_ITEM(__pyx_t_2, 1, __pyx_n_s_delayed);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_joblib, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Parallel, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_delayed); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_delayed, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0023: import multiprocessing as mp
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_multiprocessing, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 23, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_mp, __pyx_t_1) < 0) __PYX_ERR(0, 23, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0024: cimport numpy as np
+
+0025: import numpy as np
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 25, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 25, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0026: from packaging import version
+
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 26, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_version);
+  __Pyx_GIVEREF(__pyx_n_s_version);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_version);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_packaging, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 26, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_version); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 26, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_version, __pyx_t_1) < 0) __PYX_ERR(0, 26, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0027: import pandas as pd
+
  __pyx_t_2 = __Pyx_Import(__pyx_n_s_pandas, 0, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 27, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_pd, __pyx_t_2) < 0) __PYX_ERR(0, 27, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0028: import scipy.stats as stats
+
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 28, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s__14);
+  __Pyx_GIVEREF(__pyx_n_s__14);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s__14);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_scipy_stats, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 28, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_stats, __pyx_t_1) < 0) __PYX_ERR(0, 28, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0029: import sklearn
+
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_sklearn, __pyx_t_1) < 0) __PYX_ERR(0, 29, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0030: from sklearn.utils import check_array, check_random_state, check_X_y
+
  __pyx_t_1 = PyList_New(3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_INCREF(__pyx_n_s_check_array);
+  __Pyx_GIVEREF(__pyx_n_s_check_array);
+  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_check_array);
+  __Pyx_INCREF(__pyx_n_s_check_random_state);
+  __Pyx_GIVEREF(__pyx_n_s_check_random_state);
+  PyList_SET_ITEM(__pyx_t_1, 1, __pyx_n_s_check_random_state);
+  __Pyx_INCREF(__pyx_n_s_check_X_y);
+  __Pyx_GIVEREF(__pyx_n_s_check_X_y);
+  PyList_SET_ITEM(__pyx_t_1, 2, __pyx_n_s_check_X_y);
+  __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_utils, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_array); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_array, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_random_state, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_X_y, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0031: from typing import List
+
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 31, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_n_s_List);
+  __Pyx_GIVEREF(__pyx_n_s_List);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_List);
+  __pyx_t_1 = __Pyx_Import(__pyx_n_s_typing, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 31, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_List); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 31, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_List, __pyx_t_2) < 0) __PYX_ERR(0, 31, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0032: if version.parse(sklearn.__version__) >= version.parse('0.22.0'):
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_version); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_parse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_sklearn); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_version_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_version); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_parse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_tuple__15, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyObject_RichCompare(__pyx_t_1, __pyx_t_3, Py_GE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__pyx_t_4) {
+/* … */
+    goto __pyx_L2;
+  }
+/* … */
+  __pyx_tuple__15 = PyTuple_Pack(1, __pyx_kp_s_0_22_0); if (unlikely(!__pyx_tuple__15)) __PYX_ERR(0, 32, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__15);
+  __Pyx_GIVEREF(__pyx_tuple__15);
+
+0033:     from sklearn.utils._testing import ignore_warnings
+
    __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 33, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_n_s_ignore_warnings);
+    __Pyx_GIVEREF(__pyx_n_s_ignore_warnings);
+    PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_ignore_warnings);
+    __pyx_t_3 = __Pyx_Import(__pyx_n_s_sklearn_utils__testing, __pyx_t_2, -1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 33, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_3, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 33, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (PyDict_SetItem(__pyx_d, __pyx_n_s_ignore_warnings, __pyx_t_2) < 0) __PYX_ERR(0, 33, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
 0034: else:
+
+0035:     from sklearn.utils.testing import ignore_warnings
+
  /*else*/ {
+    __pyx_t_3 = PyList_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 35, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_INCREF(__pyx_n_s_ignore_warnings);
+    __Pyx_GIVEREF(__pyx_n_s_ignore_warnings);
+    PyList_SET_ITEM(__pyx_t_3, 0, __pyx_n_s_ignore_warnings);
+    __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_utils_testing, __pyx_t_3, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 35, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 35, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    if (PyDict_SetItem(__pyx_d, __pyx_n_s_ignore_warnings, __pyx_t_3) < 0) __PYX_ERR(0, 35, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  }
+  __pyx_L2:;
+
 0036:
+
 0037:
+
+0038: MAX_INT = np.iinfo(np.int32).max
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_iinfo); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_int32); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_max); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_MAX_INT, __pyx_t_1) < 0) __PYX_ERR(0, 38, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0039:
+
 0040:
+
 0041: cdef extern from "math.h":
+
 0042:     double log(double x) nogil
+
 0043:
+
 0044:
+
 0045: @cython.cfunc
+
+0046: def kl_divergence(pk: cython.float, qk: cython.float) -> cython.float:
+
static float __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(float __pyx_v_pk, float __pyx_v_qk) {
+  float __pyx_v_eps;
+  float __pyx_v_S;
+  float __pyx_r;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("kl_divergence", 0);
+/* … */
+  /* function exit code */
+  __pyx_L0:;
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
 0047:     '''
+
 0048:     Calculate KL Divergence for binary classification.
+
 0049:
+
 0050:     sum(np.array(pk) * np.log(np.array(pk) / np.array(qk)))
+
 0051:
+
 0052:     Args
+
 0053:     ----
+
 0054:     pk : float
+
 0055:         The probability of 1 in one distribution.
+
 0056:     qk : float
+
 0057:         The probability of 1 in the other distribution.
+
 0058:
+
 0059:     Returns
+
 0060:     -------
+
 0061:     S : float
+
 0062:         The KL divergence.
+
 0063:     '''
+
 0064:
+
+0065:     eps: cython.float = 1e-6
+
  __pyx_v_eps = 1e-6;
+
 0066:     S: cython.float
+
 0067:
+
+0068:     if qk == 0.:
+
  __pyx_t_1 = ((__pyx_v_qk == 0.) != 0);
+  if (__pyx_t_1) {
+/* … */
+  }
+
+0069:         return 0.
+
    __pyx_r = 0.;
+    goto __pyx_L0;
+
 0070:
+
+0071:     qk = min(max(qk, eps), 1 - eps)
+
  __pyx_t_2 = (1.0 - __pyx_v_eps);
+  __pyx_t_3 = __pyx_v_eps;
+  __pyx_t_4 = __pyx_v_qk;
+  if (((__pyx_t_3 > __pyx_t_4) != 0)) {
+    __pyx_t_5 = __pyx_t_3;
+  } else {
+    __pyx_t_5 = __pyx_t_4;
+  }
+  __pyx_t_3 = __pyx_t_5;
+  if (((__pyx_t_2 < __pyx_t_3) != 0)) {
+    __pyx_t_5 = __pyx_t_2;
+  } else {
+    __pyx_t_5 = __pyx_t_3;
+  }
+  __pyx_v_qk = __pyx_t_5;
+
 0072:
+
+0073:     if pk == 0.:
+
  __pyx_t_1 = ((__pyx_v_pk == 0.) != 0);
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L4;
+  }
+
+0074:         S = -log(1 - qk)
+
    __pyx_v_S = (-log((1.0 - __pyx_v_qk)));
+
+0075:     elif pk == 1.:
+
  __pyx_t_1 = ((__pyx_v_pk == 1.) != 0);
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L4;
+  }
+
+0076:         S = -log(qk)
+
    __pyx_v_S = (-log(__pyx_v_qk));
+
 0077:     else:
+
+0078:         S = pk * log(pk / qk) + (1 - pk) * log((1 - pk) / (1 - qk))
+
  /*else*/ {
+    __pyx_v_S = ((__pyx_v_pk * log((__pyx_v_pk / __pyx_v_qk))) + ((1.0 - __pyx_v_pk) * log(((1.0 - __pyx_v_pk) / (1.0 - __pyx_v_qk)))));
+  }
+  __pyx_L4:;
+
 0079:
+
+0080:     return S
+
  __pyx_r = __pyx_v_S;
+  goto __pyx_L0;
+
 0081:
+
 0082:
+
 0083: @cython.cfunc
+
+0084: def entropyH(p: cython.float, q: cython.float=-1.) -> cython.float:
+
static float __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(float __pyx_v_p, struct __pyx_opt_args_8causalml_9inference_4tree_6uplift_entropyH *__pyx_optional_args) {
+  float __pyx_v_q = __pyx_k_;
+  float __pyx_r;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("entropyH", 0);
+  if (__pyx_optional_args) {
+    if (__pyx_optional_args->__pyx_n > 0) {
+      __pyx_v_q = __pyx_optional_args->q;
+    }
+  }
+/* … */
+  /* function exit code */
+  __pyx_L0:;
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_k_ = (-1.);
+/* … */
+struct __pyx_opt_args_8causalml_9inference_4tree_6uplift_entropyH {
+  int __pyx_n;
+  float q;
+};
+
 0085:     '''
+
 0086:     Entropy
+
 0087:
+
 0088:     Entropy calculation for normalization.
+
 0089:
+
 0090:     Args
+
 0091:     ----
+
 0092:     p : float
+
 0093:         The probability used in the entropy calculation.
+
 0094:
+
 0095:     q : float, optional, (default = -1.)
+
 0096:         The second probability used in the entropy calculation.
+
 0097:
+
 0098:     Returns
+
 0099:     -------
+
 0100:     entropy : float
+
 0101:     '''
+
 0102:
+
+0103:     if q == -1. and p > 0.:
+
  __pyx_t_2 = ((__pyx_v_q == -1.) != 0);
+  if (__pyx_t_2) {
+  } else {
+    __pyx_t_1 = __pyx_t_2;
+    goto __pyx_L4_bool_binop_done;
+  }
+  __pyx_t_2 = ((__pyx_v_p > 0.) != 0);
+  __pyx_t_1 = __pyx_t_2;
+  __pyx_L4_bool_binop_done:;
+  if (__pyx_t_1) {
+/* … */
+  }
+
+0104:         return -p * log(p)
+
    __pyx_r = ((-__pyx_v_p) * log(__pyx_v_p));
+    goto __pyx_L0;
+
+0105:     elif q > 0.:
+
  __pyx_t_1 = ((__pyx_v_q > 0.) != 0);
+  if (__pyx_t_1) {
+/* … */
+  }
+
+0106:         return -p * log(q)
+
    __pyx_r = ((-__pyx_v_p) * log(__pyx_v_q));
+    goto __pyx_L0;
+
 0107:     else:
+
+0108:         return 0.
+
  /*else*/ {
+    __pyx_r = 0.;
+    goto __pyx_L0;
+  }
+
 0109:
+
 0110:
+
+0111: class DecisionTree:
+
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_DecisionTree, __pyx_n_s_DecisionTree, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Tree_Node_Class_Tree_node_class); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 111, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_DecisionTree, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 111, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DecisionTree, __pyx_t_2) < 0) __PYX_ERR(0, 111, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0112:     """ Tree Node Class
+
 0113:
+
 0114:     Tree node class to contain all the statistics of the tree node.
+
 0115:
+
 0116:     Parameters
+
 0117:     ----------
+
 0118:     classes_ : list of str
+
 0119:         A list of the control and treatment group names.
+
 0120:
+
 0121:     col : int, optional (default = -1)
+
 0122:         The column index for splitting the tree node to children nodes.
+
 0123:
+
 0124:     value : float, optional (default = None)
+
 0125:         The value of the feature column to split the tree node to children nodes.
+
 0126:
+
 0127:     trueBranch : object of DecisionTree
+
 0128:         The true branch tree node (feature > value).
+
 0129:
+
 0130:     falseBranch : object of DecisionTree
+
 0131:         The false branch tree node (feature > value).
+
 0132:
+
 0133:     results : list of float
+
 0134:         The classification probability P(Y=1|T) for each of the control and treatment groups
+
 0135:         in the tree node.
+
 0136:
+
 0137:     summary : list of list
+
 0138:         Summary statistics of the tree nodes, including impurity, sample size, uplift score, etc.
+
 0139:
+
 0140:     maxDiffTreatment : int
+
 0141:         The treatment index generating the maximum difference between the treatment and control groups.
+
 0142:
+
 0143:     maxDiffSign : float
+
 0144:         The sign of the maximum difference (1. or -1.).
+
 0145:
+
 0146:     nodeSummary : list of list
+
 0147:         Summary statistics of the tree nodes [P(Y=1|T), N(T)], where y_mean stands for the target metric mean
+
 0148:         and n is the sample size.
+
 0149:
+
 0150:     backupResults : list of float
+
 0151:         The positive probabilities in each of the control and treatment groups in the parent node. The parent node
+
 0152:         information is served as a backup for the children node, in case no valid statistics can be calculated from the
+
 0153:         children node, the parent node information will be used in certain cases.
+
 0154:
+
 0155:     bestTreatment : int
+
 0156:         The treatment index providing the best uplift (treatment effect).
+
 0157:
+
 0158:     upliftScore : list
+
 0159:         The uplift score of this node: [max_Diff, p_value], where max_Diff stands for the maximum treatment effect, and
+
 0160:         p_value stands for the p_value of the treatment effect.
+
 0161:
+
 0162:     matchScore : float
+
 0163:         The uplift score by filling a trained tree with validation dataset or testing dataset.
+
 0164:
+
 0165:     """
+
 0166:
+
+0167:     def __init__(self, classes_, col=-1, value=None, trueBranch=None, falseBranch=None, results=None, summary=None,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__, METH_VARARGS|METH_KEYWORDS, 0};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_classes_ = 0;
+  PyObject *__pyx_v_col = 0;
+  PyObject *__pyx_v_value = 0;
+  PyObject *__pyx_v_trueBranch = 0;
+  PyObject *__pyx_v_falseBranch = 0;
+  PyObject *__pyx_v_results = 0;
+  PyObject *__pyx_v_summary = 0;
+  PyObject *__pyx_v_maxDiffTreatment = 0;
+  PyObject *__pyx_v_maxDiffSign = 0;
+  PyObject *__pyx_v_nodeSummary = 0;
+  PyObject *__pyx_v_backupResults = 0;
+  PyObject *__pyx_v_bestTreatment = 0;
+  PyObject *__pyx_v_upliftScore = 0;
+  PyObject *__pyx_v_matchScore = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_classes,&__pyx_n_s_col,&__pyx_n_s_value,&__pyx_n_s_trueBranch,&__pyx_n_s_falseBranch,&__pyx_n_s_results,&__pyx_n_s_summary,&__pyx_n_s_maxDiffTreatment,&__pyx_n_s_maxDiffSign,&__pyx_n_s_nodeSummary,&__pyx_n_s_backupResults,&__pyx_n_s_bestTreatment,&__pyx_n_s_upliftScore,&__pyx_n_s_matchScore,0};
+    PyObject* values[15] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
+    values[2] = ((PyObject *)((PyObject *)__pyx_int_neg_1));
+    values[3] = ((PyObject *)((PyObject *)Py_None));
+    values[4] = ((PyObject *)((PyObject *)Py_None));
+    values[5] = ((PyObject *)((PyObject *)Py_None));
+    values[6] = ((PyObject *)((PyObject *)Py_None));
+    values[7] = ((PyObject *)((PyObject *)Py_None));
+/* … */
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_12DecisionTree___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_classes_, PyObject *__pyx_v_col, PyObject *__pyx_v_value, PyObject *__pyx_v_trueBranch, PyObject *__pyx_v_falseBranch, PyObject *__pyx_v_results, PyObject *__pyx_v_summary, PyObject *__pyx_v_maxDiffTreatment, PyObject *__pyx_v_maxDiffSign, PyObject *__pyx_v_nodeSummary, PyObject *__pyx_v_backupResults, PyObject *__pyx_v_bestTreatment, PyObject *__pyx_v_upliftScore, PyObject *__pyx_v_matchScore) {
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__", 0);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.DecisionTree.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__16 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_classes, __pyx_n_s_col, __pyx_n_s_value, __pyx_n_s_trueBranch, __pyx_n_s_falseBranch, __pyx_n_s_results, __pyx_n_s_summary, __pyx_n_s_maxDiffTreatment, __pyx_n_s_maxDiffSign, __pyx_n_s_nodeSummary, __pyx_n_s_backupResults, __pyx_n_s_bestTreatment, __pyx_n_s_upliftScore, __pyx_n_s_matchScore); if (unlikely(!__pyx_tuple__16)) __PYX_ERR(0, 167, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__16);
+  __Pyx_GIVEREF(__pyx_tuple__16);
+  __pyx_codeobj__17 = (PyObject*)__Pyx_PyCode_New(15, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__16, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 167, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__17)) __PYX_ERR(0, 167, __pyx_L1_error)
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__, 0, __pyx_n_s_DecisionTree___init, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__17)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 167, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__18);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_2) < 0) __PYX_ERR(0, 167, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_tuple__18 = PyTuple_Pack(13, ((PyObject *)__pyx_int_neg_1), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject*)__pyx_float_1_), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__18)) __PYX_ERR(0, 167, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__18);
+  __Pyx_GIVEREF(__pyx_tuple__18);
+
+0168:                   maxDiffTreatment=None, maxDiffSign=1., nodeSummary=None, backupResults=None, bestTreatment=None,
+
    values[8] = ((PyObject *)((PyObject *)Py_None));
+    values[9] = ((PyObject *)((PyObject*)__pyx_float_1_));
+    values[10] = ((PyObject *)((PyObject *)Py_None));
+    values[11] = ((PyObject *)((PyObject *)Py_None));
+    values[12] = ((PyObject *)((PyObject *)Py_None));
+
+0169:                   upliftScore=None, matchScore=None):
+
    values[13] = ((PyObject *)((PyObject *)Py_None));
+    values[14] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case 15: values[14] = PyTuple_GET_ITEM(__pyx_args, 14);
+        CYTHON_FALLTHROUGH;
+        case 14: values[13] = PyTuple_GET_ITEM(__pyx_args, 13);
+        CYTHON_FALLTHROUGH;
+        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
+        CYTHON_FALLTHROUGH;
+        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
+        CYTHON_FALLTHROUGH;
+        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
+        CYTHON_FALLTHROUGH;
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_classes)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 15, 1); __PYX_ERR(0, 167, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_col);
+          if (value) { values[2] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_value);
+          if (value) { values[3] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_trueBranch);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_falseBranch);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_results);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_summary);
+          if (value) { values[7] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  8:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_maxDiffTreatment);
+          if (value) { values[8] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  9:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_maxDiffSign);
+          if (value) { values[9] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 10:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_nodeSummary);
+          if (value) { values[10] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 11:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_backupResults);
+          if (value) { values[11] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 12:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_bestTreatment);
+          if (value) { values[12] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 13:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_upliftScore);
+          if (value) { values[13] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 14:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_matchScore);
+          if (value) { values[14] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 167, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case 15: values[14] = PyTuple_GET_ITEM(__pyx_args, 14);
+        CYTHON_FALLTHROUGH;
+        case 14: values[13] = PyTuple_GET_ITEM(__pyx_args, 13);
+        CYTHON_FALLTHROUGH;
+        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
+        CYTHON_FALLTHROUGH;
+        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
+        CYTHON_FALLTHROUGH;
+        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
+        CYTHON_FALLTHROUGH;
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_classes_ = values[1];
+    __pyx_v_col = values[2];
+    __pyx_v_value = values[3];
+    __pyx_v_trueBranch = values[4];
+    __pyx_v_falseBranch = values[5];
+    __pyx_v_results = values[6];
+    __pyx_v_summary = values[7];
+    __pyx_v_maxDiffTreatment = values[8];
+    __pyx_v_maxDiffSign = values[9];
+    __pyx_v_nodeSummary = values[10];
+    __pyx_v_backupResults = values[11];
+    __pyx_v_bestTreatment = values[12];
+    __pyx_v_upliftScore = values[13];
+    __pyx_v_matchScore = values[14];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 15, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 167, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.DecisionTree.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_12DecisionTree___init__(__pyx_self, __pyx_v_self, __pyx_v_classes_, __pyx_v_col, __pyx_v_value, __pyx_v_trueBranch, __pyx_v_falseBranch, __pyx_v_results, __pyx_v_summary, __pyx_v_maxDiffTreatment, __pyx_v_maxDiffSign, __pyx_v_nodeSummary, __pyx_v_backupResults, __pyx_v_bestTreatment, __pyx_v_upliftScore, __pyx_v_matchScore);
+
+0170:         self.classes_ = classes_
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_v_classes_) < 0) __PYX_ERR(0, 170, __pyx_L1_error)
+
+0171:         self.col = col
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_col, __pyx_v_col) < 0) __PYX_ERR(0, 171, __pyx_L1_error)
+
+0172:         self.value = value
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_value, __pyx_v_value) < 0) __PYX_ERR(0, 172, __pyx_L1_error)
+
+0173:         self.trueBranch = trueBranch
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_trueBranch, __pyx_v_trueBranch) < 0) __PYX_ERR(0, 173, __pyx_L1_error)
+
+0174:         self.falseBranch = falseBranch
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_falseBranch, __pyx_v_falseBranch) < 0) __PYX_ERR(0, 174, __pyx_L1_error)
+
+0175:         self.results = results  # None for nodes, not None for leaves
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_results, __pyx_v_results) < 0) __PYX_ERR(0, 175, __pyx_L1_error)
+
+0176:         self.summary = summary
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_summary, __pyx_v_summary) < 0) __PYX_ERR(0, 176, __pyx_L1_error)
+
 0177:         # the treatment with max( |p(y|treatment) - p(y|control)| )
+
+0178:         self.maxDiffTreatment = maxDiffTreatment
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 178, __pyx_L1_error)
+
 0179:         # the sign for p(y|maxDiffTreatment) - p(y|control)
+
+0180:         self.maxDiffSign = maxDiffSign
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 180, __pyx_L1_error)
+
+0181:         self.nodeSummary = nodeSummary
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_nodeSummary, __pyx_v_nodeSummary) < 0) __PYX_ERR(0, 181, __pyx_L1_error)
+
+0182:         self.backupResults = backupResults
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_backupResults, __pyx_v_backupResults) < 0) __PYX_ERR(0, 182, __pyx_L1_error)
+
+0183:         self.bestTreatment = bestTreatment
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 183, __pyx_L1_error)
+
+0184:         self.upliftScore = upliftScore
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 184, __pyx_L1_error)
+
 0185:         # match actual treatment for validation and testing
+
+0186:         self.matchScore = matchScore
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_matchScore, __pyx_v_matchScore) < 0) __PYX_ERR(0, 186, __pyx_L1_error)
+
 0187:
+
 0188:
+
 0189: # Uplift Tree Classifier
+
+0190: class UpliftTreeClassifier:
+
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_UpliftTreeClassifier, __pyx_n_s_UpliftTreeClassifier, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Uplift_Tree_Classifier_for_Clas); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 190, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_UpliftTreeClassifier, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 190, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_UpliftTreeClassifier, __pyx_t_2) < 0) __PYX_ERR(0, 190, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0191:     """ Uplift Tree Classifier for Classification Task.
+
 0192:
+
 0193:     A uplift tree classifier estimates the individual treatment effect by modifying the loss function in the
+
 0194:     classification trees.
+
 0195:
+
 0196:     The uplift tree classifier is used in uplift random forest to construct the trees in the forest.
+
 0197:
+
 0198:     Parameters
+
 0199:     ----------
+
 0200:
+
 0201:     evaluationFunction : string
+
 0202:         Choose from one of the models: 'KL', 'ED', 'Chi', 'CTS', 'DDP'.
+
 0203:
+
 0204:     max_features: int, optional (default=None)
+
 0205:         The number of features to consider when looking for the best split.
+
 0206:
+
 0207:     max_depth: int, optional (default=3)
+
 0208:         The maximum depth of the tree.
+
 0209:
+
 0210:     min_samples_leaf: int, optional (default=100)
+
 0211:         The minimum number of samples required to be split at a leaf node.
+
 0212:
+
 0213:     min_samples_treatment: int, optional (default=10)
+
 0214:         The minimum number of samples required of the experiment group to be split at a leaf node.
+
 0215:
+
 0216:     n_reg: int, optional (default=100)
+
 0217:         The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the
+
 0218:         parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
+
 0219:
+
 0220:     control_name: string
+
 0221:         The name of the control group (other experiment groups will be regarded as treatment groups).
+
 0222:
+
 0223:     normalization: boolean, optional (default=True)
+
 0224:         The normalization factor defined in Rzepakowski et al. 2012, correcting for tests with large number of splits
+
 0225:         and imbalanced treatment and control splits.
+
 0226:
+
 0227:     random_state: int, RandomState instance or None (default=None)
+
 0228:         A random seed or `np.random.RandomState` to control randomness in building a tree.
+
 0229:
+
 0230:     """
+
+0231:     def __init__(self, control_name, max_features=None, max_depth=3, min_samples_leaf=100,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__, METH_VARARGS|METH_KEYWORDS, 0};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_control_name = 0;
+  PyObject *__pyx_v_max_features = 0;
+  PyObject *__pyx_v_max_depth = 0;
+  PyObject *__pyx_v_min_samples_leaf = 0;
+  PyObject *__pyx_v_min_samples_treatment = 0;
+  PyObject *__pyx_v_n_reg = 0;
+  PyObject *__pyx_v_evaluationFunction = 0;
+  PyObject *__pyx_v_normalization = 0;
+  PyObject *__pyx_v_random_state = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_control_name,&__pyx_n_s_max_features,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_evaluationFunction,&__pyx_n_s_normalization,&__pyx_n_s_random_state,0};
+    PyObject* values[10] = {0,0,0,0,0,0,0,0,0,0};
+    values[2] = ((PyObject *)((PyObject *)Py_None));
+    values[3] = ((PyObject *)((PyObject *)__pyx_int_3));
+    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
+    values[5] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[6] = ((PyObject *)((PyObject *)__pyx_int_100));
+    values[7] = ((PyObject *)((PyObject*)__pyx_n_s_KL));
+/* … */
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_control_name, PyObject *__pyx_v_max_features, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_evaluationFunction, PyObject *__pyx_v_normalization, PyObject *__pyx_v_random_state) {
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__", 0);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__19 = PyTuple_Pack(10, __pyx_n_s_self, __pyx_n_s_control_name, __pyx_n_s_max_features, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_evaluationFunction, __pyx_n_s_normalization, __pyx_n_s_random_state); if (unlikely(!__pyx_tuple__19)) __PYX_ERR(0, 231, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__19);
+  __Pyx_GIVEREF(__pyx_tuple__19);
+  __pyx_codeobj__20 = (PyObject*)__Pyx_PyCode_New(10, 0, 10, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__19, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 231, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__20)) __PYX_ERR(0, 231, __pyx_L1_error)
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__, 0, __pyx_n_s_UpliftTreeClassifier___init, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__20)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 231, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__21);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_2) < 0) __PYX_ERR(0, 231, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_tuple__21 = PyTuple_Pack(8, ((PyObject *)Py_None), ((PyObject *)__pyx_int_3), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject*)__pyx_n_s_KL), ((PyObject *)Py_True), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__21)) __PYX_ERR(0, 231, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__21);
+  __Pyx_GIVEREF(__pyx_tuple__21);
+
 0232:                  min_samples_treatment=10, n_reg=100, evaluationFunction='KL',
+
+0233:                  normalization=True, random_state=None):
+
    values[8] = ((PyObject *)((PyObject *)Py_True));
+    values[9] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 10, 1); __PYX_ERR(0, 231, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_features);
+          if (value) { values[2] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
+          if (value) { values[3] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_evaluationFunction);
+          if (value) { values[7] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  8:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_normalization);
+          if (value) { values[8] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  9:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
+          if (value) { values[9] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 231, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_control_name = values[1];
+    __pyx_v_max_features = values[2];
+    __pyx_v_max_depth = values[3];
+    __pyx_v_min_samples_leaf = values[4];
+    __pyx_v_min_samples_treatment = values[5];
+    __pyx_v_n_reg = values[6];
+    __pyx_v_evaluationFunction = values[7];
+    __pyx_v_normalization = values[8];
+    __pyx_v_random_state = values[9];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 10, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 231, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier___init__(__pyx_self, __pyx_v_self, __pyx_v_control_name, __pyx_v_max_features, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_evaluationFunction, __pyx_v_normalization, __pyx_v_random_state);
+
+0234:         self.max_depth = max_depth
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 234, __pyx_L1_error)
+
+0235:         self.min_samples_leaf = min_samples_leaf
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 235, __pyx_L1_error)
+
+0236:         self.min_samples_treatment = min_samples_treatment
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 236, __pyx_L1_error)
+
+0237:         self.n_reg = n_reg
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 237, __pyx_L1_error)
+
+0238:         self.max_features = max_features
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_features, __pyx_v_max_features) < 0) __PYX_ERR(0, 238, __pyx_L1_error)
+
+0239:         if evaluationFunction == 'KL':
+
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_KL, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 239, __pyx_L1_error)
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L3;
+  }
+
+0240:             self.evaluationFunction = self.evaluate_KL
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_KL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 240, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 240, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0241:         elif evaluationFunction == 'ED':
+
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_ED, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 241, __pyx_L1_error)
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L3;
+  }
+
+0242:             self.evaluationFunction = self.evaluate_ED
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_ED); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 242, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 242, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0243:         elif evaluationFunction == 'Chi':
+
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_Chi, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 243, __pyx_L1_error)
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L3;
+  }
+
+0244:             self.evaluationFunction = self.evaluate_Chi
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_Chi); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 244, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 244, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0245:         elif evaluationFunction == 'DDP':
+
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_DDP, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 245, __pyx_L1_error)
+  if (__pyx_t_1) {
+/* … */
+    goto __pyx_L3;
+  }
+
+0246:             self.evaluationFunction = self.evaluate_DDP
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 246, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 246, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0247:         else:
+
+0248:             self.evaluationFunction = self.evaluate_CTS
+
  /*else*/ {
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 248, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 248, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  }
+  __pyx_L3:;
+
+0249:         self.fitted_uplift_tree = None
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree, Py_None) < 0) __PYX_ERR(0, 249, __pyx_L1_error)
+
 0250:
+
+0251:         assert control_name is not None and isinstance(control_name, str), \
+
  #ifndef CYTHON_WITHOUT_ASSERTIONS
+  if (unlikely(!Py_OptimizeFlag)) {
+    __pyx_t_3 = (__pyx_v_control_name != Py_None);
+    __pyx_t_4 = (__pyx_t_3 != 0);
+    if (__pyx_t_4) {
+    } else {
+      __pyx_t_1 = __pyx_t_4;
+      goto __pyx_L4_bool_binop_done;
+    }
+    __pyx_t_4 = PyString_Check(__pyx_v_control_name); 
+    __pyx_t_3 = (__pyx_t_4 != 0);
+    __pyx_t_1 = __pyx_t_3;
+    __pyx_L4_bool_binop_done:;
+    if (unlikely(!__pyx_t_1)) {
+
+0252:             f"control_group should be string but {control_name} is passed"
+
      __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 252, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_5 = 0;
+      __pyx_t_6 = 127;
+      __Pyx_INCREF(__pyx_kp_u_control_group_should_be_string_b);
+      __pyx_t_5 += 35;
+      __Pyx_GIVEREF(__pyx_kp_u_control_group_should_be_string_b);
+      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_kp_u_control_group_should_be_string_b);
+      __pyx_t_7 = __Pyx_PyObject_FormatSimple(__pyx_v_control_name, __pyx_empty_unicode); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 252, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __pyx_t_6 = (__Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) > __pyx_t_6) ? __Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) : __pyx_t_6;
+      __pyx_t_5 += __Pyx_PyUnicode_GET_LENGTH(__pyx_t_7);
+      __Pyx_GIVEREF(__pyx_t_7);
+      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_7);
+      __pyx_t_7 = 0;
+      __Pyx_INCREF(__pyx_kp_u_is_passed);
+      __pyx_t_5 += 10;
+      __Pyx_GIVEREF(__pyx_kp_u_is_passed);
+      PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_kp_u_is_passed);
+      __pyx_t_7 = __Pyx_PyUnicode_Join(__pyx_t_2, 3, __pyx_t_5, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 252, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      PyErr_SetObject(PyExc_AssertionError, __pyx_t_7);
+      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      __PYX_ERR(0, 251, __pyx_L1_error)
+    }
+  }
+  #endif
+
+0253:         self.control_name = control_name
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
+
+0254:         self.classes_ = [self.control_name]
+
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GIVEREF(__pyx_t_7);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_7);
+  __pyx_t_7 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 254, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0255:         self.n_class = 1
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_int_1) < 0) __PYX_ERR(0, 255, __pyx_L1_error)
+
 0256:
+
+0257:         self.normalization = normalization
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_normalization, __pyx_v_normalization) < 0) __PYX_ERR(0, 257, __pyx_L1_error)
+
+0258:         self.random_state = random_state
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 258, __pyx_L1_error)
+
 0259:
+
+0260:     def fit(self, X, treatment, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit[] = " Fit the uplift model.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        self : object\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 260, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 260, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 260, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 260, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 260, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_treatment_groups = NULL;
+  PyObject *__pyx_v_treatment_idx = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_v_col = NULL;
+  PyObject *__pyx_v_imp = NULL;
+  PyObject *__pyx_v_x = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit", 0);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment);
+  __Pyx_INCREF(__pyx_v_y);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_treatment_groups);
+  __Pyx_XDECREF(__pyx_v_treatment_idx);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XDECREF(__pyx_v_col);
+  __Pyx_XDECREF(__pyx_v_imp);
+  __Pyx_XDECREF(__pyx_v_x);
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XDECREF(__pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_v_y);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__22 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_treatment_groups, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr, __pyx_n_s_col, __pyx_n_s_imp, __pyx_n_s_x); if (unlikely(!__pyx_tuple__22)) __PYX_ERR(0, 260, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__22);
+  __Pyx_GIVEREF(__pyx_tuple__22);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit, 0, __pyx_n_s_UpliftTreeClassifier_fit, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__23)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 260, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fit, __pyx_t_2) < 0) __PYX_ERR(0, 260, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_codeobj__23 = (PyObject*)__Pyx_PyCode_New(4, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__22, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fit, 260, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__23)) __PYX_ERR(0, 260, __pyx_L1_error)
+
 0261:         """ Fit the uplift model.
+
 0262:
+
 0263:         Args
+
 0264:         ----
+
 0265:         X : ndarray, shape = [num_samples, num_features]
+
 0266:             An ndarray of the covariates used to train the uplift model.
+
 0267:
+
 0268:         treatment : array-like, shape = [num_samples]
+
 0269:             An array containing the treatment group for each unit.
+
 0270:
+
 0271:         y : array-like, shape = [num_samples]
+
 0272:             An array containing the outcome of interest for each unit.
+
 0273:
+
 0274:         Returns
+
 0275:         -------
+
 0276:         self : object
+
 0277:         """
+
 0278:
+
+0279:         self.random_state_ = check_random_state(self.random_state)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 279, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 279, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 279, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state_2, __pyx_t_1) < 0) __PYX_ERR(0, 279, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0280:
+
+0281:         X, y = check_X_y(X, y)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 281, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_5 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_5 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_4 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_X);
+    __Pyx_GIVEREF(__pyx_v_X);
+    PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_5, __pyx_v_X);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_5, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
+    PyObject* sequence = __pyx_t_1;
+    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+    if (unlikely(size != 2)) {
+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+      __PYX_ERR(0, 281, __pyx_L1_error)
+    }
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    if (likely(PyTuple_CheckExact(sequence))) {
+      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
+      __pyx_t_4 = PyTuple_GET_ITEM(sequence, 1); 
+    } else {
+      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
+      __pyx_t_4 = PyList_GET_ITEM(sequence, 1); 
+    }
+    __Pyx_INCREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_t_4);
+    #else
+    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    #endif
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  } else {
+    Py_ssize_t index = -1;
+    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
+    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_2);
+    index = 1; __pyx_t_4 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_4)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_4);
+    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 281, __pyx_L1_error)
+    __pyx_t_6 = NULL;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    goto __pyx_L4_unpacking_done;
+    __pyx_L3_unpacking_failed:;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_6 = NULL;
+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+    __PYX_ERR(0, 281, __pyx_L1_error)
+    __pyx_L4_unpacking_done:;
+  }
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
+  __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_4);
+  __pyx_t_4 = 0;
+
+0282:         treatment = np.asarray(treatment)
+
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 282, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 282, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 282, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
+  __pyx_t_1 = 0;
+
+0283:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
+
  #ifndef CYTHON_WITHOUT_ASSERTIONS
+  if (unlikely(!Py_OptimizeFlag)) {
+    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 283, __pyx_L1_error)
+    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 283, __pyx_L1_error)
+    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
+      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
+      __PYX_ERR(0, 283, __pyx_L1_error)
+    }
+  }
+  #endif
+
 0284:
+
 0285:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
+
+0286:         treatment_groups = sorted([x for x in list(set(treatment)) if x != self.control_name])
+
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 286, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = PySet_New(__pyx_v_treatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 286, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_3 = PySequence_List(__pyx_t_4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = __pyx_t_3; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  for (;;) {
+    if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 286, __pyx_L1_error)
+    #else
+    __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    #endif
+    __Pyx_XDECREF_SET(__pyx_v_x, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_9 = PyObject_RichCompare(__pyx_v_x, __pyx_t_3, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 286, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 286, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    if (__pyx_t_10) {
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_v_x))) __PYX_ERR(0, 286, __pyx_L1_error)
+    }
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_1 = ((PyObject*)__pyx_t_2);
+  __pyx_t_2 = 0;
+  __pyx_t_11 = PyList_Sort(__pyx_t_1); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 286, __pyx_L1_error)
+  __pyx_v_treatment_groups = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
+0287:         self.classes_ = [self.control_name]
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 287, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 287, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GIVEREF(__pyx_t_1);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
+  __pyx_t_1 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 287, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0288:         treatment_idx = np.zeros_like(treatment, dtype=int)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_treatment);
+  __Pyx_GIVEREF(__pyx_v_treatment);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment);
+  __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_dtype, ((PyObject *)(&PyInt_Type))) < 0) __PYX_ERR(0, 288, __pyx_L1_error)
+  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_v_treatment_idx = __pyx_t_9;
+  __pyx_t_9 = 0;
+
+0289:         for i, tr in enumerate(treatment_groups, 1):
+
  __Pyx_INCREF(__pyx_int_1);
+  __pyx_t_9 = __pyx_int_1;
+  __pyx_t_4 = __pyx_v_treatment_groups; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
+  for (;;) {
+    if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    __pyx_t_2 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 289, __pyx_L1_error)
+    #else
+    __pyx_t_2 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    #endif
+    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
+    __pyx_t_2 = 0;
+    __Pyx_INCREF(__pyx_t_9);
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_9);
+    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_9, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_9);
+    __pyx_t_9 = __pyx_t_2;
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+
+0290:             self.classes_.append(tr)
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 290, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_11 = __Pyx_PyObject_Append(__pyx_t_2, __pyx_v_tr); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 290, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0291:             treatment_idx[treatment == tr] = i
+
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
+    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 291, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0292:         self.n_class = len(self.classes_)
+
  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 292, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_8 = PyObject_Length(__pyx_t_9); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 292, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = PyInt_FromSsize_t(__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 292, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_t_9) < 0) __PYX_ERR(0, 292, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+
 0293:
+
+0294:         self.feature_imp_dict = defaultdict(float)
+
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 294, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_4, function);
+    }
+  }
+  __pyx_t_9 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_2, ((PyObject *)(&PyFloat_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_4, ((PyObject *)(&PyFloat_Type)));
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 294, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict, __pyx_t_9) < 0) __PYX_ERR(0, 294, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+
 0295:
+
+0296:         if self.evaluationFunction == self.evaluate_DDP and self.n_class > 2:
+
  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = PyObject_RichCompare(__pyx_t_9, __pyx_t_4, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__pyx_t_12) {
+  } else {
+    __pyx_t_10 = __pyx_t_12;
+    goto __pyx_L11_bool_binop_done;
+  }
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = PyObject_RichCompare(__pyx_t_2, __pyx_int_2, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 296, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_10 = __pyx_t_12;
+  __pyx_L11_bool_binop_done:;
+  if (unlikely(__pyx_t_10)) {
+/* … */
+  }
+
+0297:             raise ValueError("The DDP approach can only cope with two class problems, that is two different treatment "
+
    __pyx_t_4 = __Pyx_PyObject_Call(__pyx_builtin_ValueError, __pyx_tuple__2, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 297, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_Raise(__pyx_t_4, 0, 0, 0);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __PYX_ERR(0, 297, __pyx_L1_error)
+/* … */
+  __pyx_tuple__2 = PyTuple_Pack(1, __pyx_kp_s_The_DDP_approach_can_only_cope_w); if (unlikely(!__pyx_tuple__2)) __PYX_ERR(0, 297, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__2);
+  __Pyx_GIVEREF(__pyx_tuple__2);
+
 0298:                              "options (e.g., control vs treatment). Please select another approach or only use a "
+
 0299:                              "dataset which employs two treatment options.")
+
 0300:
+
+0301:         self.fitted_uplift_tree = self.growDecisionTreeFrom(
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 301, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+/* … */
+  __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 301, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_v_X);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_y);
+/* … */
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_2, __pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 301, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree, __pyx_t_1) < 0) __PYX_ERR(0, 301, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 0302:             X, treatment_idx, y,
+
+0303:             max_depth=self.max_depth, min_samples_leaf=self.min_samples_leaf,
+
  __pyx_t_9 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_max_depth, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_leaf, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_depth, __pyx_int_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+
+0304:             depth=1, min_samples_treatment=self.min_samples_treatment,
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 304, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0305:             n_reg=self.n_reg, parentNodeSummary=None
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 305, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
+
 0306:         )
+
 0307:
+
+0308:         self.feature_importances_ = np.zeros(X.shape[1])
+
  __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_zeros); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_9, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_9)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_9);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_9, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_1) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0309:         for col, imp in self.feature_imp_dict.items():
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 309, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_items); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 309, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_4, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_2) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
+    __pyx_t_4 = __pyx_t_1; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
+    __pyx_t_13 = NULL;
+  } else {
+    __pyx_t_8 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 309, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_13 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 309, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_13)) {
+      if (likely(PyList_CheckExact(__pyx_t_4))) {
+        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_1); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 309, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      } else {
+        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_1); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 309, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      }
+    } else {
+      __pyx_t_1 = __pyx_t_13(__pyx_t_4);
+      if (unlikely(!__pyx_t_1)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 309, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_1);
+    }
+    if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
+      PyObject* sequence = __pyx_t_1;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 2)) {
+        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 309, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_9 = PyTuple_GET_ITEM(sequence, 1); 
+      } else {
+        __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_9 = PyList_GET_ITEM(sequence, 1); 
+      }
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_9);
+      #else
+      __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 309, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_9 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 309, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      #endif
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 309, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
+      index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L15_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_2);
+      index = 1; __pyx_t_9 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_9)) goto __pyx_L15_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_9);
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 309, __pyx_L1_error)
+      __pyx_t_6 = NULL;
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      goto __pyx_L16_unpacking_done;
+      __pyx_L15_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_6 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 309, __pyx_L1_error)
+      __pyx_L16_unpacking_done:;
+    }
+    __Pyx_XDECREF_SET(__pyx_v_col, __pyx_t_2);
+    __pyx_t_2 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_imp, __pyx_t_9);
+    __pyx_t_9 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0310:             self.feature_importances_[col] = imp
+
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 310, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_v_col, __pyx_v_imp) < 0)) __PYX_ERR(0, 310, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0311:         self.feature_importances_ /= self.feature_importances_.sum()  # normalize to add to 1
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_sum); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_9)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_9);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_9) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyNumber_InPlaceDivide(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_2) < 0) __PYX_ERR(0, 311, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0312:
+
 0313:     # Prune Trees
+
+0314:     def prune(self, X, treatment, y, minGain=0.0001, rule='maxAbsDiff'):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune[] = " Prune the uplift model.\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        minGain : float, optional (default = 0.0001)\n            The minimum gain required to make a tree node split. The children\n            tree branches are trimmed if the actual split gain is less than\n            the minimum gain.\n        rule : string, optional (default = 'maxAbsDiff')\n            The prune rules. Supported values are 'maxAbsDiff' for optimizing\n            the maximum absolute difference, and 'bestUplift' for optimizing\n            the node-size weighted treatment effect.\n        Returns\n        -------\n        self : object\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune = {"prune", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_minGain = 0;
+  PyObject *__pyx_v_rule = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("prune (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_minGain,&__pyx_n_s_rule,0};
+    PyObject* values[6] = {0,0,0,0,0,0};
+    values[4] = ((PyObject *)((PyObject*)__pyx_float_0_0001));
+    values[5] = ((PyObject *)((PyObject*)__pyx_n_s_maxAbsDiff));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 1); __PYX_ERR(0, 314, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 2); __PYX_ERR(0, 314, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 3); __PYX_ERR(0, 314, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_minGain);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_rule);
+          if (value) { values[5] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "prune") < 0)) __PYX_ERR(0, 314, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_minGain = values[4];
+    __pyx_v_rule = values[5];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 314, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.prune", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_minGain, __pyx_v_rule);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_minGain, PyObject *__pyx_v_rule) {
+  PyObject *__pyx_v_treatment_idx = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("prune", 0);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment);
+  __Pyx_INCREF(__pyx_v_y);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.prune", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_treatment_idx);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XDECREF(__pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_v_y);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__24 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_minGain, __pyx_n_s_rule, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__24)) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__24);
+  __Pyx_GIVEREF(__pyx_tuple__24);
+  __pyx_codeobj__25 = (PyObject*)__Pyx_PyCode_New(6, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__24, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_prune, 314, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__25)) __PYX_ERR(0, 314, __pyx_L1_error)
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune, 0, __pyx_n_s_UpliftTreeClassifier_prune, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__25)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__26);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_prune, __pyx_t_2) < 0) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_tuple__26 = PyTuple_Pack(2, ((PyObject*)__pyx_float_0_0001), ((PyObject*)__pyx_n_s_maxAbsDiff)); if (unlikely(!__pyx_tuple__26)) __PYX_ERR(0, 314, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__26);
+  __Pyx_GIVEREF(__pyx_tuple__26);
+
 0315:         """ Prune the uplift model.
+
 0316:         Args
+
 0317:         ----
+
 0318:         X : ndarray, shape = [num_samples, num_features]
+
 0319:             An ndarray of the covariates used to train the uplift model.
+
 0320:         treatment : array-like, shape = [num_samples]
+
 0321:             An array containing the treatment group for each unit.
+
 0322:         y : array-like, shape = [num_samples]
+
 0323:             An array containing the outcome of interest for each unit.
+
 0324:         minGain : float, optional (default = 0.0001)
+
 0325:             The minimum gain required to make a tree node split. The children
+
 0326:             tree branches are trimmed if the actual split gain is less than
+
 0327:             the minimum gain.
+
 0328:         rule : string, optional (default = 'maxAbsDiff')
+
 0329:             The prune rules. Supported values are 'maxAbsDiff' for optimizing
+
 0330:             the maximum absolute difference, and 'bestUplift' for optimizing
+
 0331:             the node-size weighted treatment effect.
+
 0332:         Returns
+
 0333:         -------
+
 0334:         self : object
+
 0335:         """
+
 0336:
+
+0337:         X, y = check_X_y(X, y)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 337, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_X);
+    __Pyx_GIVEREF(__pyx_v_X);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
+    PyObject* sequence = __pyx_t_1;
+    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+    if (unlikely(size != 2)) {
+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+      __PYX_ERR(0, 337, __pyx_L1_error)
+    }
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    if (likely(PyTuple_CheckExact(sequence))) {
+      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
+      __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
+    } else {
+      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
+      __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
+    }
+    __Pyx_INCREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_t_5);
+    #else
+    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    #endif
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  } else {
+    Py_ssize_t index = -1;
+    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 337, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
+    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_2);
+    index = 1; __pyx_t_5 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_5)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 337, __pyx_L1_error)
+    __pyx_t_6 = NULL;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    goto __pyx_L4_unpacking_done;
+    __pyx_L3_unpacking_failed:;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_6 = NULL;
+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+    __PYX_ERR(0, 337, __pyx_L1_error)
+    __pyx_L4_unpacking_done:;
+  }
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
+  __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_5);
+  __pyx_t_5 = 0;
+
+0338:         treatment = np.asarray(treatment)
+
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 338, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 338, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 338, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
+  __pyx_t_1 = 0;
+
+0339:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
+
  #ifndef CYTHON_WITHOUT_ASSERTIONS
+  if (unlikely(!Py_OptimizeFlag)) {
+    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 339, __pyx_L1_error)
+    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 339, __pyx_L1_error)
+    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
+      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
+      __PYX_ERR(0, 339, __pyx_L1_error)
+    }
+  }
+  #endif
+
 0340:
+
 0341:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
+
+0342:         treatment_idx = np.zeros_like(treatment)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_5);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_5, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_2, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 342, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_treatment_idx = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+0343:         for i, tr in enumerate(self.classes_[1:], 1):
+
  __Pyx_INCREF(__pyx_int_1);
+  __pyx_t_1 = __pyx_int_1;
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 343, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_t_5, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_8 = 0;
+    __pyx_t_9 = NULL;
+  } else {
+    __pyx_t_8 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 343, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_9 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 343, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_9)) {
+      if (likely(PyList_CheckExact(__pyx_t_5))) {
+        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_5)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 343, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 343, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_9(__pyx_t_5);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 343, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
+    __pyx_t_2 = 0;
+    __Pyx_INCREF(__pyx_t_1);
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
+    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1);
+    __pyx_t_1 = __pyx_t_2;
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+/* … */
+  __pyx_slice__3 = PySlice_New(__pyx_int_1, Py_None, Py_None); if (unlikely(!__pyx_slice__3)) __PYX_ERR(0, 343, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_slice__3);
+  __Pyx_GIVEREF(__pyx_slice__3);
+
+0344:             treatment_idx[treatment == tr] = i
+
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 344, __pyx_L1_error)
+    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 344, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0345:
+
+0346:         self.pruneTree(X, treatment_idx, y,
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 346, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_5 = PyTuple_New(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 346, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_v_X);
+  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_5, 2, __pyx_v_y);
+/* … */
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 346, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0347:                        tree=self.fitted_uplift_tree,
+
  __pyx_t_2 = __Pyx_PyDict_NewPresized(5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 347, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 347, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_tree, __pyx_t_3) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0348:                        rule=rule,
+
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_rule, __pyx_v_rule) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
+
+0349:                        minGain=minGain,
+
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_minGain, __pyx_v_minGain) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
+
+0350:                        n_reg=self.n_reg,
+
  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 350, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_reg, __pyx_t_3) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0351:                        parentNodeSummary=None)
+
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
+
+0352:         return self
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_self);
+  __pyx_r = __pyx_v_self;
+  goto __pyx_L0;
+
 0353:
+
+0354:     def pruneTree(self, X, treatment_idx, y, tree, rule='maxAbsDiff', minGain=0.,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree[] = "Prune one single tree node in the uplift model.\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        rule : string, optional (default = 'maxAbsDiff')\n            The prune rules. Supported values are 'maxAbsDiff' for optimizing the maximum absolute difference, and\n            'bestUplift' for optimizing the node-size weighted treatment effect.\n        minGain : float, optional (default = 0.)\n            The minimum gain required to make a tree node split. The children tree branches are trimmed if the actual\n            split gain is less than the minimum gain.\n        n_reg: int, optional (default=0)\n            The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the\n            parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : list of list, optional (default = None)\n            Node summary statistics, [P(Y=1|T), N(T)] of the parent tree node.\n        Returns\n        -------\n        self : object\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree = {"pruneTree", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_v_rule = 0;
+  PyObject *__pyx_v_minGain = 0;
+  PyObject *__pyx_v_n_reg = 0;
+  PyObject *__pyx_v_parentNodeSummary = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("pruneTree (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_tree,&__pyx_n_s_rule,&__pyx_n_s_minGain,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
+    PyObject* values[9] = {0,0,0,0,0,0,0,0,0};
+    values[5] = ((PyObject *)((PyObject*)__pyx_n_s_maxAbsDiff));
+    values[6] = ((PyObject *)((PyObject*)__pyx_float_0_));
+    values[7] = ((PyObject *)((PyObject *)__pyx_int_0));
+/* … */
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_tree, PyObject *__pyx_v_rule, PyObject *__pyx_v_minGain, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
+  PyObject *__pyx_v_currentNodeSummary = NULL;
+  PyObject *__pyx_v_X_l = NULL;
+  PyObject *__pyx_v_X_r = NULL;
+  PyObject *__pyx_v_w_l = NULL;
+  PyObject *__pyx_v_w_r = NULL;
+  PyObject *__pyx_v_y_l = NULL;
+  PyObject *__pyx_v_y_r = NULL;
+  PyObject *__pyx_v_currentScoreD = NULL;
+  PyObject *__pyx_v_trueNodeSummary = NULL;
+  PyObject *__pyx_v_trueScoreD = NULL;
+  PyObject *__pyx_v_falseNodeSummary = NULL;
+  PyObject *__pyx_v_falseScoreD = NULL;
+  PyObject *__pyx_v_gain = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("pruneTree", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_XDECREF(__pyx_t_12);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.pruneTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
+  __Pyx_XDECREF(__pyx_v_X_l);
+  __Pyx_XDECREF(__pyx_v_X_r);
+  __Pyx_XDECREF(__pyx_v_w_l);
+  __Pyx_XDECREF(__pyx_v_w_r);
+  __Pyx_XDECREF(__pyx_v_y_l);
+  __Pyx_XDECREF(__pyx_v_y_r);
+  __Pyx_XDECREF(__pyx_v_currentScoreD);
+  __Pyx_XDECREF(__pyx_v_trueNodeSummary);
+  __Pyx_XDECREF(__pyx_v_trueScoreD);
+  __Pyx_XDECREF(__pyx_v_falseNodeSummary);
+  __Pyx_XDECREF(__pyx_v_falseScoreD);
+  __Pyx_XDECREF(__pyx_v_gain);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__27 = PyTuple_Pack(22, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_rule, __pyx_n_s_minGain, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_currentNodeSummary, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_currentScoreD, __pyx_n_s_trueNodeSummary, __pyx_n_s_trueScoreD, __pyx_n_s_falseNodeSummary, __pyx_n_s_falseScoreD, __pyx_n_s_gain); if (unlikely(!__pyx_tuple__27)) __PYX_ERR(0, 354, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__27);
+  __Pyx_GIVEREF(__pyx_tuple__27);
+  __pyx_codeobj__28 = (PyObject*)__Pyx_PyCode_New(9, 0, 22, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__27, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_pruneTree, 354, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__28)) __PYX_ERR(0, 354, __pyx_L1_error)
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree, 0, __pyx_n_s_UpliftTreeClassifier_pruneTree, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__28)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 354, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__29);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_pruneTree, __pyx_t_2) < 0) __PYX_ERR(0, 354, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_tuple__29 = PyTuple_Pack(4, ((PyObject*)__pyx_n_s_maxAbsDiff), ((PyObject*)__pyx_float_0_), ((PyObject *)__pyx_int_0), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__29)) __PYX_ERR(0, 354, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__29);
+  __Pyx_GIVEREF(__pyx_tuple__29);
+
 0355:                   n_reg=0,
+
+0356:                   parentNodeSummary=None):
+
    values[8] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 1); __PYX_ERR(0, 354, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 2); __PYX_ERR(0, 354, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 3); __PYX_ERR(0, 354, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 4); __PYX_ERR(0, 354, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_rule);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_minGain);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
+          if (value) { values[7] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  8:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
+          if (value) { values[8] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "pruneTree") < 0)) __PYX_ERR(0, 354, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment_idx = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_tree = values[4];
+    __pyx_v_rule = values[5];
+    __pyx_v_minGain = values[6];
+    __pyx_v_n_reg = values[7];
+    __pyx_v_parentNodeSummary = values[8];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 354, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.pruneTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_tree, __pyx_v_rule, __pyx_v_minGain, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
+
 0357:         """Prune one single tree node in the uplift model.
+
 0358:         Args
+
 0359:         ----
+
 0360:         X : ndarray, shape = [num_samples, num_features]
+
 0361:             An ndarray of the covariates used to train the uplift model.
+
 0362:         treatment_idx : array-like, shape = [num_samples]
+
 0363:             An array containing the treatment group index for each unit.
+
 0364:         y : array-like, shape = [num_samples]
+
 0365:             An array containing the outcome of interest for each unit.
+
 0366:         rule : string, optional (default = 'maxAbsDiff')
+
 0367:             The prune rules. Supported values are 'maxAbsDiff' for optimizing the maximum absolute difference, and
+
 0368:             'bestUplift' for optimizing the node-size weighted treatment effect.
+
 0369:         minGain : float, optional (default = 0.)
+
 0370:             The minimum gain required to make a tree node split. The children tree branches are trimmed if the actual
+
 0371:             split gain is less than the minimum gain.
+
 0372:         n_reg: int, optional (default=0)
+
 0373:             The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the
+
 0374:             parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
+
 0375:         parentNodeSummary : list of list, optional (default = None)
+
 0376:             Node summary statistics, [P(Y=1|T), N(T)] of the parent tree node.
+
 0377:         Returns
+
 0378:         -------
+
 0379:         self : object
+
 0380:         """
+
 0381:         # Current Node Summary for Validation Data Set
+
+0382:         currentNodeSummary = self.tree_node_summary(
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 382, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 382, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_y);
+/* … */
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 382, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_v_currentNodeSummary = __pyx_t_4;
+  __pyx_t_4 = 0;
+
+0383:             treatment_idx, y, min_samples_treatment=self.min_samples_treatment,
+
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 383, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 383, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_t_4) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0384:             n_reg=n_reg, parentNodeSummary=parentNodeSummary
+
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, __pyx_v_parentNodeSummary) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
+
 0385:         )
+
+0386:         tree.nodeSummary = currentNodeSummary
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 386, __pyx_L1_error)
+
 0387:         # Divide sets for child nodes
+
+0388:         if (tree.trueBranch is None) or (tree.falseBranch is None):
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 388, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_6 = (__pyx_t_4 == Py_None);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_7 = (__pyx_t_6 != 0);
+  if (!__pyx_t_7) {
+  } else {
+    __pyx_t_5 = __pyx_t_7;
+    goto __pyx_L4_bool_binop_done;
+  }
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 388, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_7 = (__pyx_t_4 == Py_None);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_6 = (__pyx_t_7 != 0);
+  __pyx_t_5 = __pyx_t_6;
+  __pyx_L4_bool_binop_done:;
+  if (__pyx_t_5) {
+/* … */
+  }
+
+0389:             X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, tree.col, tree.value)
+
    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_8 = NULL;
+    __pyx_t_9 = 0;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
+      __pyx_t_8 = PyMethod_GET_SELF(__pyx_t_3);
+      if (likely(__pyx_t_8)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+        __Pyx_INCREF(__pyx_t_8);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_3, function);
+        __pyx_t_9 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_3)) {
+      PyObject *__pyx_temp[6] = {__pyx_t_8, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
+      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_9, 5+__pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
+      PyObject *__pyx_temp[6] = {__pyx_t_8, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
+      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_9, 5+__pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    } else
+    #endif
+    {
+      __pyx_t_10 = PyTuple_New(5+__pyx_t_9); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 389, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      if (__pyx_t_8) {
+        __Pyx_GIVEREF(__pyx_t_8); PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_t_8); __pyx_t_8 = NULL;
+      }
+      __Pyx_INCREF(__pyx_v_X);
+      __Pyx_GIVEREF(__pyx_v_X);
+      PyTuple_SET_ITEM(__pyx_t_10, 0+__pyx_t_9, __pyx_v_X);
+      __Pyx_INCREF(__pyx_v_treatment_idx);
+      __Pyx_GIVEREF(__pyx_v_treatment_idx);
+      PyTuple_SET_ITEM(__pyx_t_10, 1+__pyx_t_9, __pyx_v_treatment_idx);
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_10, 2+__pyx_t_9, __pyx_v_y);
+      __Pyx_GIVEREF(__pyx_t_2);
+      PyTuple_SET_ITEM(__pyx_t_10, 3+__pyx_t_9, __pyx_t_2);
+      __Pyx_GIVEREF(__pyx_t_1);
+      PyTuple_SET_ITEM(__pyx_t_10, 4+__pyx_t_9, __pyx_t_1);
+      __pyx_t_2 = 0;
+      __pyx_t_1 = 0;
+      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_10, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
+      PyObject* sequence = __pyx_t_4;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 6)) {
+        if (size > 6) __Pyx_RaiseTooManyValuesError(6);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 389, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_10 = PyTuple_GET_ITEM(sequence, 1); 
+        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 2); 
+        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 3); 
+        __pyx_t_8 = PyTuple_GET_ITEM(sequence, 4); 
+        __pyx_t_11 = PyTuple_GET_ITEM(sequence, 5); 
+      } else {
+        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_10 = PyList_GET_ITEM(sequence, 1); 
+        __pyx_t_1 = PyList_GET_ITEM(sequence, 2); 
+        __pyx_t_2 = PyList_GET_ITEM(sequence, 3); 
+        __pyx_t_8 = PyList_GET_ITEM(sequence, 4); 
+        __pyx_t_11 = PyList_GET_ITEM(sequence, 5); 
+      }
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_10);
+      __Pyx_INCREF(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_8);
+      __Pyx_INCREF(__pyx_t_11);
+      #else
+      {
+        Py_ssize_t i;
+        PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_10,&__pyx_t_1,&__pyx_t_2,&__pyx_t_8,&__pyx_t_11};
+        for (i=0; i < 6; i++) {
+          PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 389, __pyx_L1_error)
+          __Pyx_GOTREF(item);
+          *(temps[i]) = item;
+        }
+      }
+      #endif
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_10,&__pyx_t_1,&__pyx_t_2,&__pyx_t_8,&__pyx_t_11};
+      __pyx_t_12 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 389, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_13 = Py_TYPE(__pyx_t_12)->tp_iternext;
+      for (index=0; index < 6; index++) {
+        PyObject* item = __pyx_t_13(__pyx_t_12); if (unlikely(!item)) goto __pyx_L6_unpacking_failed;
+        __Pyx_GOTREF(item);
+        *(temps[index]) = item;
+      }
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_13(__pyx_t_12), 6) < 0) __PYX_ERR(0, 389, __pyx_L1_error)
+      __pyx_t_13 = NULL;
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      goto __pyx_L7_unpacking_done;
+      __pyx_L6_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      __pyx_t_13 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 389, __pyx_L1_error)
+      __pyx_L7_unpacking_done:;
+    }
+    __pyx_v_X_l = __pyx_t_3;
+    __pyx_t_3 = 0;
+    __pyx_v_X_r = __pyx_t_10;
+    __pyx_t_10 = 0;
+    __pyx_v_w_l = __pyx_t_1;
+    __pyx_t_1 = 0;
+    __pyx_v_w_r = __pyx_t_2;
+    __pyx_t_2 = 0;
+    __pyx_v_y_l = __pyx_t_8;
+    __pyx_t_8 = 0;
+    __pyx_v_y_r = __pyx_t_11;
+    __pyx_t_11 = 0;
+
 0390:
+
 0391:             # recursive call for each branch
+
+0392:             if tree.trueBranch.results is None:
+
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_results); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_11);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_5 = (__pyx_t_11 == Py_None);
+    __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+    __pyx_t_6 = (__pyx_t_5 != 0);
+    if (__pyx_t_6) {
+/* … */
+    }
+
+0393:                 self.pruneTree(X_l, w_l, y_l, tree.trueBranch, rule, minGain,
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 393, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 393, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+/* … */
+      __pyx_t_8 = PyTuple_New(7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 393, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_INCREF(__pyx_v_X_l);
+      __Pyx_GIVEREF(__pyx_v_X_l);
+      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_X_l);
+      __Pyx_INCREF(__pyx_v_w_l);
+      __Pyx_GIVEREF(__pyx_v_w_l);
+      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_w_l);
+      __Pyx_INCREF(__pyx_v_y_l);
+      __Pyx_GIVEREF(__pyx_v_y_l);
+      PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_v_y_l);
+      __Pyx_GIVEREF(__pyx_t_4);
+      PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_4);
+      __Pyx_INCREF(__pyx_v_rule);
+      __Pyx_GIVEREF(__pyx_v_rule);
+      PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_v_rule);
+      __Pyx_INCREF(__pyx_v_minGain);
+      __Pyx_GIVEREF(__pyx_v_minGain);
+      PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_v_minGain);
+      __Pyx_INCREF(__pyx_v_n_reg);
+      __Pyx_GIVEREF(__pyx_v_n_reg);
+      PyTuple_SET_ITEM(__pyx_t_8, 6, __pyx_v_n_reg);
+      __pyx_t_4 = 0;
+/* … */
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_8, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 393, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0394:                                n_reg,
+
+0395:                                parentNodeSummary=currentNodeSummary)
+
      __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 395, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 395, __pyx_L1_error)
+
+0396:             if tree.falseBranch.results is None:
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 396, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_results); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 396, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_6 = (__pyx_t_4 == Py_None);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_5 = (__pyx_t_6 != 0);
+    if (__pyx_t_5) {
+/* … */
+    }
+
+0397:                 self.pruneTree(X_r, w_r, y_r, tree.falseBranch, rule, minGain,
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 397, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 397, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+/* … */
+      __pyx_t_8 = PyTuple_New(7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 397, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_INCREF(__pyx_v_X_r);
+      __Pyx_GIVEREF(__pyx_v_X_r);
+      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_X_r);
+      __Pyx_INCREF(__pyx_v_w_r);
+      __Pyx_GIVEREF(__pyx_v_w_r);
+      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_w_r);
+      __Pyx_INCREF(__pyx_v_y_r);
+      __Pyx_GIVEREF(__pyx_v_y_r);
+      PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_v_y_r);
+      __Pyx_GIVEREF(__pyx_t_2);
+      PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_2);
+      __Pyx_INCREF(__pyx_v_rule);
+      __Pyx_GIVEREF(__pyx_v_rule);
+      PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_v_rule);
+      __Pyx_INCREF(__pyx_v_minGain);
+      __Pyx_GIVEREF(__pyx_v_minGain);
+      PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_v_minGain);
+      __Pyx_INCREF(__pyx_v_n_reg);
+      __Pyx_GIVEREF(__pyx_v_n_reg);
+      PyTuple_SET_ITEM(__pyx_t_8, 6, __pyx_v_n_reg);
+      __pyx_t_2 = 0;
+/* … */
+      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 397, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
 0398:                                n_reg,
+
+0399:                                parentNodeSummary=currentNodeSummary)
+
      __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 399, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 399, __pyx_L1_error)
+
 0400:
+
 0401:         # merge leaves (potentially)
+
+0402:         if (tree.trueBranch.results is not None and
+
  __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_results); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __pyx_t_6 = (__pyx_t_2 != Py_None);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_7 = (__pyx_t_6 != 0);
+  if (__pyx_t_7) {
+  } else {
+    __pyx_t_5 = __pyx_t_7;
+    goto __pyx_L11_bool_binop_done;
+  }
+/* … */
+  if (__pyx_t_5) {
+/* … */
+  }
+
+0403:             tree.falseBranch.results is not None):
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 403, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_results); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 403, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_7 = (__pyx_t_11 != Py_None);
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __pyx_t_6 = (__pyx_t_7 != 0);
+  __pyx_t_5 = __pyx_t_6;
+  __pyx_L11_bool_binop_done:;
+
+0404:             if rule == 'maxAbsDiff':
+
    __pyx_t_5 = (__Pyx_PyString_Equals(__pyx_v_rule, __pyx_n_s_maxAbsDiff, Py_EQ)); if (unlikely(__pyx_t_5 < 0)) __PYX_ERR(0, 404, __pyx_L1_error)
+    if (__pyx_t_5) {
+/* … */
+      goto __pyx_L13;
+    }
+
 0405:                 # Current D
+
+0406:                 if (tree.maxDiffTreatment in currentNodeSummary and
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 406, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 406, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L15_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L14;
+      }
+
+0407:                     self.control_name in currentNodeSummary):
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 407, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 407, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L15_bool_binop_done:;
+
+0408:                     currentScoreD = tree.maxDiffSign * (currentNodeSummary[tree.maxDiffTreatment][0]
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 408, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 408, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 408, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 408, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+/* … */
+        __pyx_t_8 = PyNumber_Multiply(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 408, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_v_currentScoreD = __pyx_t_8;
+        __pyx_t_8 = 0;
+
+0409:                                                         - currentNodeSummary[self.control_name][0])
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 409, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 409, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 409, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = PyNumber_Subtract(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 409, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
 0410:                 else:
+
+0411:                     currentScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_currentScoreD = __pyx_int_0;
+      }
+      __pyx_L14:;
+
 0412:
+
 0413:                 # trueBranch D
+
+0414:                 trueNodeSummary = self.tree_node_summary(
+
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 414, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+/* … */
+      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 414, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_INCREF(__pyx_v_w_l);
+      __Pyx_GIVEREF(__pyx_v_w_l);
+      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_l);
+      __Pyx_INCREF(__pyx_v_y_l);
+      __Pyx_GIVEREF(__pyx_v_y_l);
+      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_l);
+/* … */
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 414, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_v_trueNodeSummary = __pyx_t_2;
+      __pyx_t_2 = 0;
+
+0415:                     w_l, y_l, min_samples_treatment=self.min_samples_treatment,
+
      if (unlikely(!__pyx_v_w_l)) { __Pyx_RaiseUnboundLocalError("w_l"); __PYX_ERR(0, 415, __pyx_L1_error) }
+      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 415, __pyx_L1_error) }
+/* … */
+      __pyx_t_11 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 415, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 415, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_min_samples_treatment, __pyx_t_2) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0416:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
+
 0417:                 )
+
+0418:                 if (tree.trueBranch.maxDiffTreatment in trueNodeSummary and
+
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 418, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 418, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 418, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L18_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L17;
+      }
+
+0419:                     self.control_name in trueNodeSummary):
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 419, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 419, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L18_bool_binop_done:;
+
+0420:                     trueScoreD = tree.trueBranch.maxDiffSign * (trueNodeSummary[tree.trueBranch.maxDiffTreatment][0]
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+/* … */
+        __pyx_t_11 = PyNumber_Multiply(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_v_trueScoreD = __pyx_t_11;
+        __pyx_t_11 = 0;
+
+0421:                                                                 - trueNodeSummary[self.control_name][0])
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 421, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 421, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 421, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = PyNumber_Subtract(__pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 421, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
 0422:                     trueScoreD = (
+
 0423:                         trueScoreD
+
+0424:                         * (trueNodeSummary[tree.trueBranch.maxDiffTreatment][1]
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 424, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_11, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 424, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+/* … */
+        __pyx_t_11 = PyNumber_Multiply(__pyx_v_trueScoreD, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0425:                          + trueNodeSummary[self.control_name][1])
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 425, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 425, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 425, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = PyNumber_Add(__pyx_t_8, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 425, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0426:                         / (currentNodeSummary[tree.trueBranch.maxDiffTreatment][1]
+
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 426, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_2, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 426, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+/* … */
+        __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_DECREF_SET(__pyx_v_trueScoreD, __pyx_t_2);
+        __pyx_t_2 = 0;
+
+0427:                            + currentNodeSummary[self.control_name][1])
+
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 427, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 427, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 427, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = PyNumber_Add(__pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 427, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0428:                     )
+
 0429:                 else:
+
+0430:                     trueScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_trueScoreD = __pyx_int_0;
+      }
+      __pyx_L17:;
+
 0431:
+
 0432:                 # falseBranch D
+
+0433:                 falseNodeSummary = self.tree_node_summary(
+
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 433, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+/* … */
+      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 433, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_INCREF(__pyx_v_w_r);
+      __Pyx_GIVEREF(__pyx_v_w_r);
+      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_r);
+      __Pyx_INCREF(__pyx_v_y_r);
+      __Pyx_GIVEREF(__pyx_v_y_r);
+      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_r);
+/* … */
+      __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 433, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_v_falseNodeSummary = __pyx_t_8;
+      __pyx_t_8 = 0;
+
+0434:                     w_r, y_r, min_samples_treatment=self.min_samples_treatment,
+
      if (unlikely(!__pyx_v_w_r)) { __Pyx_RaiseUnboundLocalError("w_r"); __PYX_ERR(0, 434, __pyx_L1_error) }
+      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 434, __pyx_L1_error) }
+/* … */
+      __pyx_t_11 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 434, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 434, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_min_samples_treatment, __pyx_t_8) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+0435:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
+
 0436:                 )
+
+0437:                 if (tree.falseBranch.maxDiffTreatment in falseNodeSummary and
+
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 437, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 437, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 437, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L21_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L20;
+      }
+
+0438:                     self.control_name in falseNodeSummary):
+
+      /* "causalml/inference/tree/uplift.pyx":438
+ *                 )
+ *                 if (tree.falseBranch.maxDiffTreatment in falseNodeSummary and
+ *                     self.control_name in falseNodeSummary):             # <<<<<<<<<<<<<<
+ *                     falseScoreD = (
+ *                         tree.falseBranch.maxDiffSign *
+ */
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 438, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 438, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L21_bool_binop_done:;
+
 0439:                     falseScoreD = (
+
+0440:                         tree.falseBranch.maxDiffSign *
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 440, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 440, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+/* … */
+        __pyx_t_11 = PyNumber_Multiply(__pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 440, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_v_falseScoreD = __pyx_t_11;
+        __pyx_t_11 = 0;
+
+0441:                         (falseNodeSummary[tree.falseBranch.maxDiffTreatment][0]
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 441, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 441, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 441, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 441, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0442:                          - falseNodeSummary[self.control_name][0])
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 442, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 442, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 442, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = PyNumber_Subtract(__pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 442, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
 0443:                     )
+
 0444:
+
 0445:                     falseScoreD = (
+
+0446:                         falseScoreD *
+
        __pyx_t_11 = PyNumber_Multiply(__pyx_v_falseScoreD, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 446, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+0447:                         (falseNodeSummary[tree.falseBranch.maxDiffTreatment][1]
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 447, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 447, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 447, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_11, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 447, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0448:                          + falseNodeSummary[self.control_name][1])
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 448, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 448, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_8, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 448, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = PyNumber_Add(__pyx_t_2, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 448, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0449:                         / (currentNodeSummary[tree.falseBranch.maxDiffTreatment][1]
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 449, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 449, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+/* … */
+        __pyx_t_8 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_DECREF_SET(__pyx_v_falseScoreD, __pyx_t_8);
+        __pyx_t_8 = 0;
+
+0450:                            + currentNodeSummary[self.control_name][1])
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 450, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 450, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 450, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = PyNumber_Add(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 450, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
 0451:                     )
+
 0452:                 else:
+
+0453:                     falseScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_falseScoreD = __pyx_int_0;
+      }
+      __pyx_L20:;
+
 0454:
+
+0455:                 if ((trueScoreD + falseScoreD) - currentScoreD <= minGain or
+
      __pyx_t_8 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 455, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_4 = PyNumber_Subtract(__pyx_t_8, __pyx_v_currentScoreD); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 455, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_8 = PyObject_RichCompare(__pyx_t_4, __pyx_v_minGain, Py_LE); __Pyx_XGOTREF(__pyx_t_8); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 455, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_8); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 455, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      if (!__pyx_t_6) {
+      } else {
+        __pyx_t_5 = __pyx_t_6;
+        goto __pyx_L24_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+      }
+
+0456:                     (trueScoreD + falseScoreD < 0.)):
+
      __pyx_t_8 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 456, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_4 = PyObject_RichCompare(__pyx_t_8, __pyx_float_0_, Py_LT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 456, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 456, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L24_bool_binop_done:;
+
+0457:                     tree.trueBranch, tree.falseBranch = None, None
+
        __pyx_t_4 = Py_None;
+        __Pyx_INCREF(__pyx_t_4);
+        __pyx_t_8 = Py_None;
+        __Pyx_INCREF(__pyx_t_8);
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch, __pyx_t_4) < 0) __PYX_ERR(0, 457, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch, __pyx_t_8) < 0) __PYX_ERR(0, 457, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+0458:                     tree.results = tree.backupResults
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_backupResults); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 458, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_8) < 0) __PYX_ERR(0, 458, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
 0459:
+
+0460:             elif rule == 'bestUplift':
+
    __pyx_t_5 = (__Pyx_PyString_Equals(__pyx_v_rule, __pyx_n_s_bestUplift, Py_EQ)); if (unlikely(__pyx_t_5 < 0)) __PYX_ERR(0, 460, __pyx_L1_error)
+    if (__pyx_t_5) {
+/* … */
+    }
+    __pyx_L13:;
+
 0461:                 # Current D
+
+0462:                 if (tree.bestTreatment in currentNodeSummary and
+
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 462, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 462, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L27_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L26;
+      }
+
+0463:                     self.control_name in currentNodeSummary):
+
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 463, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 463, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L27_bool_binop_done:;
+
 0464:                     currentScoreD = (
+
+0465:                         currentNodeSummary[tree.bestTreatment][0]
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 465, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 465, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 465, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0466:                         - currentNodeSummary[self.control_name][0]
+
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 466, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 466, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 466, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = PyNumber_Subtract(__pyx_t_8, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 466, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_v_currentScoreD = __pyx_t_11;
+        __pyx_t_11 = 0;
+
 0467:                     )
+
 0468:                 else:
+
+0469:                     currentScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_currentScoreD = __pyx_int_0;
+      }
+      __pyx_L26:;
+
 0470:
+
 0471:                 # trueBranch D
+
+0472:                 trueNodeSummary = self.tree_node_summary(
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 472, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+/* … */
+      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 472, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_INCREF(__pyx_v_w_l);
+      __Pyx_GIVEREF(__pyx_v_w_l);
+      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_l);
+      __Pyx_INCREF(__pyx_v_y_l);
+      __Pyx_GIVEREF(__pyx_v_y_l);
+      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_l);
+/* … */
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_4, __pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 472, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_v_trueNodeSummary = __pyx_t_2;
+      __pyx_t_2 = 0;
+
+0473:                     w_l, y_l, min_samples_treatment=self.min_samples_treatment,
+
      if (unlikely(!__pyx_v_w_l)) { __Pyx_RaiseUnboundLocalError("w_l"); __PYX_ERR(0, 473, __pyx_L1_error) }
+      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 473, __pyx_L1_error) }
+/* … */
+      __pyx_t_8 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 473, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 473, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_min_samples_treatment, __pyx_t_2) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0474:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
+
 0475:                 )
+
+0476:                 if (tree.trueBranch.bestTreatment in trueNodeSummary and
+
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 476, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 476, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 476, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L30_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L29;
+      }
+
+0477:                     self.control_name in trueNodeSummary):
+
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 477, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 477, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L30_bool_binop_done:;
+
 0478:                     trueScoreD = (
+
+0479:                         trueNodeSummary[tree.trueBranch.bestTreatment][0]
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 479, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 479, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 479, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 479, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+0480:                         - trueNodeSummary[self.control_name][0]
+
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 480, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 480, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 480, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_4 = PyNumber_Subtract(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 480, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_v_trueScoreD = __pyx_t_4;
+        __pyx_t_4 = 0;
+
 0481:                     )
+
 0482:                 else:
+
+0483:                     trueScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_trueScoreD = __pyx_int_0;
+      }
+      __pyx_L29:;
+
 0484:
+
 0485:                 # falseBranch D
+
+0486:                 falseNodeSummary = self.tree_node_summary(
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 486, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+/* … */
+      __pyx_t_8 = PyTuple_New(2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 486, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_INCREF(__pyx_v_w_r);
+      __Pyx_GIVEREF(__pyx_v_w_r);
+      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_w_r);
+      __Pyx_INCREF(__pyx_v_y_r);
+      __Pyx_GIVEREF(__pyx_v_y_r);
+      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_y_r);
+/* … */
+      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 486, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_v_falseNodeSummary = __pyx_t_11;
+      __pyx_t_11 = 0;
+
+0487:                     w_r, y_r, min_samples_treatment=self.min_samples_treatment,
+
      if (unlikely(!__pyx_v_w_r)) { __Pyx_RaiseUnboundLocalError("w_r"); __PYX_ERR(0, 487, __pyx_L1_error) }
+      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 487, __pyx_L1_error) }
+/* … */
+      __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 487, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 487, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_min_samples_treatment, __pyx_t_11) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0488:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
+
 0489:                 )
+
+0490:                 if (tree.falseBranch.bestTreatment in falseNodeSummary and
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 490, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 490, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_2, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 490, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_7 = (__pyx_t_6 != 0);
+      if (__pyx_t_7) {
+      } else {
+        __pyx_t_5 = __pyx_t_7;
+        goto __pyx_L33_bool_binop_done;
+      }
+/* … */
+      if (__pyx_t_5) {
+/* … */
+        goto __pyx_L32;
+      }
+
+0491:                     self.control_name in falseNodeSummary):
+
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 491, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_2, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 491, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_6 = (__pyx_t_7 != 0);
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L33_bool_binop_done:;
+
 0492:                     falseScoreD = (
+
+0493:                         falseNodeSummary[tree.falseBranch.bestTreatment][0]
+
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 493, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 493, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 493, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 493, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0494:                         - falseNodeSummary[self.control_name][0]
+
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 494, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 494, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 494, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __pyx_t_8 = PyNumber_Subtract(__pyx_t_11, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 494, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __pyx_v_falseScoreD = __pyx_t_8;
+        __pyx_t_8 = 0;
+
 0495:                     )
+
 0496:                 else:
+
+0497:                     falseScoreD = 0
+
      /*else*/ {
+        __Pyx_INCREF(__pyx_int_0);
+        __pyx_v_falseScoreD = __pyx_int_0;
+      }
+      __pyx_L32:;
+
+0498:                 gain = ((1. * len(y_l) / len(y) * trueScoreD
+
      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 498, __pyx_L1_error) }
+      __pyx_t_14 = PyObject_Length(__pyx_v_y_l); if (unlikely(__pyx_t_14 == ((Py_ssize_t)-1))) __PYX_ERR(0, 498, __pyx_L1_error)
+      __pyx_t_15 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 498, __pyx_L1_error)
+      __pyx_t_8 = PyFloat_FromDouble(((1. * __pyx_t_14) / __pyx_t_15)); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 498, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_2 = PyNumber_Multiply(__pyx_t_8, __pyx_v_trueScoreD); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 498, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+0499:                          + 1. * len(y_r) / len(y) * falseScoreD)
+
      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 499, __pyx_L1_error) }
+      __pyx_t_15 = PyObject_Length(__pyx_v_y_r); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 499, __pyx_L1_error)
+      __pyx_t_14 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_14 == ((Py_ssize_t)-1))) __PYX_ERR(0, 499, __pyx_L1_error)
+      __pyx_t_8 = PyFloat_FromDouble(((1. * __pyx_t_15) / __pyx_t_14)); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 499, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __pyx_t_11 = PyNumber_Multiply(__pyx_t_8, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 499, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_8 = PyNumber_Add(__pyx_t_2, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 499, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0500:                         - currentScoreD)
+
      __pyx_t_11 = PyNumber_Subtract(__pyx_t_8, __pyx_v_currentScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 500, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_v_gain = __pyx_t_11;
+      __pyx_t_11 = 0;
+
+0501:                 if gain <= minGain or (trueScoreD + falseScoreD < 0.):
+
      __pyx_t_11 = PyObject_RichCompare(__pyx_v_gain, __pyx_v_minGain, Py_LE); __Pyx_XGOTREF(__pyx_t_11); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 501, __pyx_L1_error)
+      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_11); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 501, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      if (!__pyx_t_6) {
+      } else {
+        __pyx_t_5 = __pyx_t_6;
+        goto __pyx_L36_bool_binop_done;
+      }
+      __pyx_t_11 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 501, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_8 = PyObject_RichCompare(__pyx_t_11, __pyx_float_0_, Py_LT); __Pyx_XGOTREF(__pyx_t_8); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 501, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_8); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 501, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+      __pyx_t_5 = __pyx_t_6;
+      __pyx_L36_bool_binop_done:;
+      if (__pyx_t_5) {
+/* … */
+      }
+
+0502:                     tree.trueBranch, tree.falseBranch = None, None
+
        __pyx_t_8 = Py_None;
+        __Pyx_INCREF(__pyx_t_8);
+        __pyx_t_11 = Py_None;
+        __Pyx_INCREF(__pyx_t_11);
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch, __pyx_t_8) < 0) __PYX_ERR(0, 502, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch, __pyx_t_11) < 0) __PYX_ERR(0, 502, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0503:                     tree.results = tree.backupResults
+
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_backupResults); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 503, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_11) < 0) __PYX_ERR(0, 503, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+
+0504:         return self
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_self);
+  __pyx_r = __pyx_v_self;
+  goto __pyx_L0;
+
 0505:
+
+0506:     def fill(self, X, treatment, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill[] = " Fill the data into an existing tree.\n        This is a higher-level function to transform the original data inputs\n        into lower level data inputs (list of list and tree).\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        self : object\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill = {"fill", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fill (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 1); __PYX_ERR(0, 506, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 2); __PYX_ERR(0, 506, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 3); __PYX_ERR(0, 506, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fill") < 0)) __PYX_ERR(0, 506, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 506, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fill", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_treatment_idx = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fill", 0);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment);
+  __Pyx_INCREF(__pyx_v_y);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fill", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_treatment_idx);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XDECREF(__pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_v_y);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__30 = PyTuple_Pack(7, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__30)) __PYX_ERR(0, 506, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__30);
+  __Pyx_GIVEREF(__pyx_tuple__30);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill, 0, __pyx_n_s_UpliftTreeClassifier_fill, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__31)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 506, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fill, __pyx_t_2) < 0) __PYX_ERR(0, 506, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_codeobj__31 = (PyObject*)__Pyx_PyCode_New(4, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__30, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fill, 506, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__31)) __PYX_ERR(0, 506, __pyx_L1_error)
+
 0507:         """ Fill the data into an existing tree.
+
 0508:         This is a higher-level function to transform the original data inputs
+
 0509:         into lower level data inputs (list of list and tree).
+
 0510:
+
 0511:         Args
+
 0512:         ----
+
 0513:         X : ndarray, shape = [num_samples, num_features]
+
 0514:             An ndarray of the covariates used to train the uplift model.
+
 0515:         treatment : array-like, shape = [num_samples]
+
 0516:             An array containing the treatment group for each unit.
+
 0517:         y : array-like, shape = [num_samples]
+
 0518:             An array containing the outcome of interest for each unit.
+
 0519:
+
 0520:         Returns
+
 0521:         -------
+
 0522:         self : object
+
 0523:         """
+
 0524:
+
+0525:         X, y = check_X_y(X, y)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 525, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_X);
+    __Pyx_GIVEREF(__pyx_v_X);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
+    PyObject* sequence = __pyx_t_1;
+    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+    if (unlikely(size != 2)) {
+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+      __PYX_ERR(0, 525, __pyx_L1_error)
+    }
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    if (likely(PyTuple_CheckExact(sequence))) {
+      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
+      __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
+    } else {
+      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
+      __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
+    }
+    __Pyx_INCREF(__pyx_t_2);
+    __Pyx_INCREF(__pyx_t_5);
+    #else
+    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    #endif
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  } else {
+    Py_ssize_t index = -1;
+    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 525, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
+    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_2);
+    index = 1; __pyx_t_5 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_5)) goto __pyx_L3_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 525, __pyx_L1_error)
+    __pyx_t_6 = NULL;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    goto __pyx_L4_unpacking_done;
+    __pyx_L3_unpacking_failed:;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_6 = NULL;
+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+    __PYX_ERR(0, 525, __pyx_L1_error)
+    __pyx_L4_unpacking_done:;
+  }
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
+  __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_5);
+  __pyx_t_5 = 0;
+
+0526:         treatment = np.asarray(treatment)
+
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 526, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 526, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 526, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
+  __pyx_t_1 = 0;
+
+0527:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
+
  #ifndef CYTHON_WITHOUT_ASSERTIONS
+  if (unlikely(!Py_OptimizeFlag)) {
+    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 527, __pyx_L1_error)
+    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 527, __pyx_L1_error)
+    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
+      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
+      __PYX_ERR(0, 527, __pyx_L1_error)
+    }
+  }
+  #endif
+
 0528:
+
 0529:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
+
+0530:         treatment_idx = np.zeros_like(treatment)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 530, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 530, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_5);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_5, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_2, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_v_treatment);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 530, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_treatment_idx = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+0531:         for i, tr in enumerate(self.classes_[1:], 1):
+
  __Pyx_INCREF(__pyx_int_1);
+  __pyx_t_1 = __pyx_int_1;
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 531, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_t_5, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_8 = 0;
+    __pyx_t_9 = NULL;
+  } else {
+    __pyx_t_8 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 531, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_9 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 531, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_9)) {
+      if (likely(PyList_CheckExact(__pyx_t_5))) {
+        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_5)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 531, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 531, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_9(__pyx_t_5);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 531, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
+    __pyx_t_2 = 0;
+    __Pyx_INCREF(__pyx_t_1);
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
+    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1);
+    __pyx_t_1 = __pyx_t_2;
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0532:             treatment_idx[treatment == tr] = i
+
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 532, __pyx_L1_error)
+    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 532, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0533:
+
+0534:         self.fillTree(X, treatment_idx, y, tree=self.fitted_uplift_tree)
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_5 = PyTuple_New(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_INCREF(__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_v_X);
+  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_X);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_5, 2, __pyx_v_y);
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_tree, __pyx_t_3) < 0) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 534, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0535:         return self
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_self);
+  __pyx_r = __pyx_v_self;
+  goto __pyx_L0;
+
 0536:
+
+0537:     def fillTree(self, X, treatment_idx, y, tree):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree[] = " Fill the data into an existing tree.\n        This is a lower-level function to execute on the tree filling task.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        tree : object\n            object of DecisionTree class\n\n        Returns\n        -------\n        self : object\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree = {"fillTree", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fillTree (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_tree,0};
+    PyObject* values[5] = {0,0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 1); __PYX_ERR(0, 537, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 2); __PYX_ERR(0, 537, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 3); __PYX_ERR(0, 537, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 4); __PYX_ERR(0, 537, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fillTree") < 0)) __PYX_ERR(0, 537, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 5) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+      values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment_idx = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_tree = values[4];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 537, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fillTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_tree);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_tree) {
+  PyObject *__pyx_v_currentNodeSummary = NULL;
+  PyObject *__pyx_v_X_l = NULL;
+  PyObject *__pyx_v_X_r = NULL;
+  PyObject *__pyx_v_w_l = NULL;
+  PyObject *__pyx_v_w_r = NULL;
+  PyObject *__pyx_v_y_l = NULL;
+  PyObject *__pyx_v_y_r = NULL;
+  PyObject *__pyx_v_matchScore = NULL;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_v_summary = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fillTree", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_XDECREF(__pyx_t_15);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fillTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
+  __Pyx_XDECREF(__pyx_v_X_l);
+  __Pyx_XDECREF(__pyx_v_X_r);
+  __Pyx_XDECREF(__pyx_v_w_l);
+  __Pyx_XDECREF(__pyx_v_w_r);
+  __Pyx_XDECREF(__pyx_v_y_l);
+  __Pyx_XDECREF(__pyx_v_y_r);
+  __Pyx_XDECREF(__pyx_v_matchScore);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XDECREF(__pyx_v_summary);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__32 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_currentNodeSummary, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_matchScore, __pyx_n_s_treatment_group, __pyx_n_s_summary); if (unlikely(!__pyx_tuple__32)) __PYX_ERR(0, 537, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__32);
+  __Pyx_GIVEREF(__pyx_tuple__32);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree, 0, __pyx_n_s_UpliftTreeClassifier_fillTree, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__33)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 537, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fillTree, __pyx_t_2) < 0) __PYX_ERR(0, 537, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_codeobj__33 = (PyObject*)__Pyx_PyCode_New(5, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__32, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fillTree, 537, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__33)) __PYX_ERR(0, 537, __pyx_L1_error)
+
 0538:         """ Fill the data into an existing tree.
+
 0539:         This is a lower-level function to execute on the tree filling task.
+
 0540:
+
 0541:         Args
+
 0542:         ----
+
 0543:         X : ndarray, shape = [num_samples, num_features]
+
 0544:             An ndarray of the covariates used to train the uplift model.
+
 0545:         treatment_idx : array-like, shape = [num_samples]
+
 0546:             An array containing the treatment group index for each unit.
+
 0547:         y : array-like, shape = [num_samples]
+
 0548:             An array containing the outcome of interest for each unit.
+
 0549:         tree : object
+
 0550:             object of DecisionTree class
+
 0551:
+
 0552:         Returns
+
 0553:         -------
+
 0554:         self : object
+
 0555:         """
+
 0556:         # Current Node Summary for Validation Data Set
+
+0557:         currentNodeSummary = self.tree_node_summary(treatment_idx, y,
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 557, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 557, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_y);
+/* … */
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 557, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_v_currentNodeSummary = __pyx_t_4;
+  __pyx_t_4 = 0;
+
+0558:                                                     min_samples_treatment=0,
+
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 558, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_int_0) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_int_0) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
+
 0559:                                                     n_reg=0,
+
+0560:                                                     parentNodeSummary=None)
+
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
+
+0561:         tree.nodeSummary = currentNodeSummary
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 561, __pyx_L1_error)
+
 0562:
+
 0563:         # Divide sets for child nodes
+
+0564:         if tree.trueBranch or tree.falseBranch:
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 564, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 564, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (!__pyx_t_6) {
+  } else {
+    __pyx_t_5 = __pyx_t_6;
+    goto __pyx_L4_bool_binop_done;
+  }
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 564, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 564, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_5 = __pyx_t_6;
+  __pyx_L4_bool_binop_done:;
+  if (__pyx_t_5) {
+/* … */
+  }
+
+0565:             X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, tree.col, tree.value)
+
    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 565, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 565, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 565, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_7 = NULL;
+    __pyx_t_8 = 0;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
+      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_3);
+      if (likely(__pyx_t_7)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+        __Pyx_INCREF(__pyx_t_7);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_3, function);
+        __pyx_t_8 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_3)) {
+      PyObject *__pyx_temp[6] = {__pyx_t_7, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
+      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 5+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
+      PyObject *__pyx_temp[6] = {__pyx_t_7, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
+      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 5+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    } else
+    #endif
+    {
+      __pyx_t_9 = PyTuple_New(5+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 565, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      if (__pyx_t_7) {
+        __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_7); __pyx_t_7 = NULL;
+      }
+      __Pyx_INCREF(__pyx_v_X);
+      __Pyx_GIVEREF(__pyx_v_X);
+      PyTuple_SET_ITEM(__pyx_t_9, 0+__pyx_t_8, __pyx_v_X);
+      __Pyx_INCREF(__pyx_v_treatment_idx);
+      __Pyx_GIVEREF(__pyx_v_treatment_idx);
+      PyTuple_SET_ITEM(__pyx_t_9, 1+__pyx_t_8, __pyx_v_treatment_idx);
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_9, 2+__pyx_t_8, __pyx_v_y);
+      __Pyx_GIVEREF(__pyx_t_2);
+      PyTuple_SET_ITEM(__pyx_t_9, 3+__pyx_t_8, __pyx_t_2);
+      __Pyx_GIVEREF(__pyx_t_1);
+      PyTuple_SET_ITEM(__pyx_t_9, 4+__pyx_t_8, __pyx_t_1);
+      __pyx_t_2 = 0;
+      __pyx_t_1 = 0;
+      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_9, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
+      PyObject* sequence = __pyx_t_4;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 6)) {
+        if (size > 6) __Pyx_RaiseTooManyValuesError(6);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 565, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_9 = PyTuple_GET_ITEM(sequence, 1); 
+        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 2); 
+        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 3); 
+        __pyx_t_7 = PyTuple_GET_ITEM(sequence, 4); 
+        __pyx_t_10 = PyTuple_GET_ITEM(sequence, 5); 
+      } else {
+        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_9 = PyList_GET_ITEM(sequence, 1); 
+        __pyx_t_1 = PyList_GET_ITEM(sequence, 2); 
+        __pyx_t_2 = PyList_GET_ITEM(sequence, 3); 
+        __pyx_t_7 = PyList_GET_ITEM(sequence, 4); 
+        __pyx_t_10 = PyList_GET_ITEM(sequence, 5); 
+      }
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_9);
+      __Pyx_INCREF(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_7);
+      __Pyx_INCREF(__pyx_t_10);
+      #else
+      {
+        Py_ssize_t i;
+        PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_9,&__pyx_t_1,&__pyx_t_2,&__pyx_t_7,&__pyx_t_10};
+        for (i=0; i < 6; i++) {
+          PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 565, __pyx_L1_error)
+          __Pyx_GOTREF(item);
+          *(temps[i]) = item;
+        }
+      }
+      #endif
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_9,&__pyx_t_1,&__pyx_t_2,&__pyx_t_7,&__pyx_t_10};
+      __pyx_t_11 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 565, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_12 = Py_TYPE(__pyx_t_11)->tp_iternext;
+      for (index=0; index < 6; index++) {
+        PyObject* item = __pyx_t_12(__pyx_t_11); if (unlikely(!item)) goto __pyx_L6_unpacking_failed;
+        __Pyx_GOTREF(item);
+        *(temps[index]) = item;
+      }
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_11), 6) < 0) __PYX_ERR(0, 565, __pyx_L1_error)
+      __pyx_t_12 = NULL;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      goto __pyx_L7_unpacking_done;
+      __pyx_L6_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __pyx_t_12 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 565, __pyx_L1_error)
+      __pyx_L7_unpacking_done:;
+    }
+    __pyx_v_X_l = __pyx_t_3;
+    __pyx_t_3 = 0;
+    __pyx_v_X_r = __pyx_t_9;
+    __pyx_t_9 = 0;
+    __pyx_v_w_l = __pyx_t_1;
+    __pyx_t_1 = 0;
+    __pyx_v_w_r = __pyx_t_2;
+    __pyx_t_2 = 0;
+    __pyx_v_y_l = __pyx_t_7;
+    __pyx_t_7 = 0;
+    __pyx_v_y_r = __pyx_t_10;
+    __pyx_t_10 = 0;
+
 0566:
+
 0567:             # recursive call for each branch
+
+0568:             if tree.trueBranch is not None:
+
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 568, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = (__pyx_t_4 != Py_None);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_6 = (__pyx_t_5 != 0);
+    if (__pyx_t_6) {
+/* … */
+    }
+
+0569:                 self.fillTree(X_l, w_l, y_l, tree.trueBranch)
+
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 569, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 569, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __pyx_t_2 = NULL;
+      __pyx_t_8 = 0;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+        __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_10);
+        if (likely(__pyx_t_2)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+          __Pyx_INCREF(__pyx_t_2);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_10, function);
+          __pyx_t_8 = 1;
+        }
+      }
+      #if CYTHON_FAST_PYCALL
+      if (PyFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[5] = {__pyx_t_2, __pyx_v_X_l, __pyx_v_w_l, __pyx_v_y_l, __pyx_t_7};
+        __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      } else
+      #endif
+      #if CYTHON_FAST_PYCCALL
+      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[5] = {__pyx_t_2, __pyx_v_X_l, __pyx_v_w_l, __pyx_v_y_l, __pyx_t_7};
+        __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      } else
+      #endif
+      {
+        __pyx_t_1 = PyTuple_New(4+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 569, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        if (__pyx_t_2) {
+          __Pyx_GIVEREF(__pyx_t_2); PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_2); __pyx_t_2 = NULL;
+        }
+        __Pyx_INCREF(__pyx_v_X_l);
+        __Pyx_GIVEREF(__pyx_v_X_l);
+        PyTuple_SET_ITEM(__pyx_t_1, 0+__pyx_t_8, __pyx_v_X_l);
+        __Pyx_INCREF(__pyx_v_w_l);
+        __Pyx_GIVEREF(__pyx_v_w_l);
+        PyTuple_SET_ITEM(__pyx_t_1, 1+__pyx_t_8, __pyx_v_w_l);
+        __Pyx_INCREF(__pyx_v_y_l);
+        __Pyx_GIVEREF(__pyx_v_y_l);
+        PyTuple_SET_ITEM(__pyx_t_1, 2+__pyx_t_8, __pyx_v_y_l);
+        __Pyx_GIVEREF(__pyx_t_7);
+        PyTuple_SET_ITEM(__pyx_t_1, 3+__pyx_t_8, __pyx_t_7);
+        __pyx_t_7 = 0;
+        __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_1, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0570:             if tree.falseBranch is not None:
+
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 570, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_6 = (__pyx_t_4 != Py_None);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_5 = (__pyx_t_6 != 0);
+    if (__pyx_t_5) {
+/* … */
+    }
+
+0571:                 self.fillTree(X_r, w_r, y_r, tree.falseBranch)
+
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 571, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 571, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_7 = NULL;
+      __pyx_t_8 = 0;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+        __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_10);
+        if (likely(__pyx_t_7)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+          __Pyx_INCREF(__pyx_t_7);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_10, function);
+          __pyx_t_8 = 1;
+        }
+      }
+      #if CYTHON_FAST_PYCALL
+      if (PyFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[5] = {__pyx_t_7, __pyx_v_X_r, __pyx_v_w_r, __pyx_v_y_r, __pyx_t_1};
+        __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      } else
+      #endif
+      #if CYTHON_FAST_PYCCALL
+      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[5] = {__pyx_t_7, __pyx_v_X_r, __pyx_v_w_r, __pyx_v_y_r, __pyx_t_1};
+        __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      } else
+      #endif
+      {
+        __pyx_t_2 = PyTuple_New(4+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 571, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        if (__pyx_t_7) {
+          __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_7); __pyx_t_7 = NULL;
+        }
+        __Pyx_INCREF(__pyx_v_X_r);
+        __Pyx_GIVEREF(__pyx_v_X_r);
+        PyTuple_SET_ITEM(__pyx_t_2, 0+__pyx_t_8, __pyx_v_X_r);
+        __Pyx_INCREF(__pyx_v_w_r);
+        __Pyx_GIVEREF(__pyx_v_w_r);
+        PyTuple_SET_ITEM(__pyx_t_2, 1+__pyx_t_8, __pyx_v_w_r);
+        __Pyx_INCREF(__pyx_v_y_r);
+        __Pyx_GIVEREF(__pyx_v_y_r);
+        PyTuple_SET_ITEM(__pyx_t_2, 2+__pyx_t_8, __pyx_v_y_r);
+        __Pyx_GIVEREF(__pyx_t_1);
+        PyTuple_SET_ITEM(__pyx_t_2, 3+__pyx_t_8, __pyx_t_1);
+        __pyx_t_1 = 0;
+        __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_2, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
 0572:
+
 0573:         # Update Information
+
 0574:
+
 0575:         # matchScore
+
+0576:         matchScore = (currentNodeSummary[tree.bestTreatment][0] - currentNodeSummary[0][0])
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_10 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = PyNumber_Subtract(__pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_matchScore = __pyx_t_10;
+  __pyx_t_10 = 0;
+
+0577:         tree.matchScore = round(matchScore, 4)
+
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 577, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_INCREF(__pyx_v_matchScore);
+  __Pyx_GIVEREF(__pyx_v_matchScore);
+  PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_v_matchScore);
+  __Pyx_INCREF(__pyx_int_4);
+  __Pyx_GIVEREF(__pyx_int_4);
+  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
+  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 577, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_matchScore, __pyx_t_2) < 0) __PYX_ERR(0, 577, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0578:         tree.summary['matchScore'] = round(matchScore, 4)
+
  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 578, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_matchScore);
+  __Pyx_GIVEREF(__pyx_v_matchScore);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_matchScore);
+  __Pyx_INCREF(__pyx_int_4);
+  __Pyx_GIVEREF(__pyx_int_4);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_int_4);
+  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_2, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 578, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 578, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_n_s_matchScore, __pyx_t_10) < 0)) __PYX_ERR(0, 578, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
 0579:
+
 0580:         # Samples, Group_size
+
+0581:         tree.summary['samples'] = len(y)
+
  __pyx_t_13 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_13 == ((Py_ssize_t)-1))) __PYX_ERR(0, 581, __pyx_L1_error)
+  __pyx_t_10 = PyInt_FromSsize_t(__pyx_t_13); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 581, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 581, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_n_s_samples, __pyx_t_10) < 0)) __PYX_ERR(0, 581, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
+0582:         tree.summary['group_size'] = ''
+
  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 582, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  if (unlikely(PyObject_SetItem(__pyx_t_10, __pyx_n_s_group_size, __pyx_kp_s__4) < 0)) __PYX_ERR(0, 582, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
+0583:         for treatment_group, summary in zip(self.classes_, currentNodeSummary):
+
  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 583, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GIVEREF(__pyx_t_10);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_10);
+  __Pyx_INCREF(__pyx_v_currentNodeSummary);
+  __Pyx_GIVEREF(__pyx_v_currentNodeSummary);
+  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_currentNodeSummary);
+  __pyx_t_10 = 0;
+  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_builtin_zip, __pyx_t_2, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_10)) || PyTuple_CheckExact(__pyx_t_10)) {
+    __pyx_t_2 = __pyx_t_10; __Pyx_INCREF(__pyx_t_2); __pyx_t_13 = 0;
+    __pyx_t_14 = NULL;
+  } else {
+    __pyx_t_13 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_10); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 583, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_14 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 583, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_14)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_13 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_10 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_13); __Pyx_INCREF(__pyx_t_10); __pyx_t_13++; if (unlikely(0 < 0)) __PYX_ERR(0, 583, __pyx_L1_error)
+        #else
+        __pyx_t_10 = PySequence_ITEM(__pyx_t_2, __pyx_t_13); __pyx_t_13++; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        #endif
+      } else {
+        if (__pyx_t_13 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_10 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_13); __Pyx_INCREF(__pyx_t_10); __pyx_t_13++; if (unlikely(0 < 0)) __PYX_ERR(0, 583, __pyx_L1_error)
+        #else
+        __pyx_t_10 = PySequence_ITEM(__pyx_t_2, __pyx_t_13); __pyx_t_13++; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_10);
+        #endif
+      }
+    } else {
+      __pyx_t_10 = __pyx_t_14(__pyx_t_2);
+      if (unlikely(!__pyx_t_10)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 583, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_10);
+    }
+    if ((likely(PyTuple_CheckExact(__pyx_t_10))) || (PyList_CheckExact(__pyx_t_10))) {
+      PyObject* sequence = __pyx_t_10;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 2)) {
+        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 583, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_4 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 1); 
+      } else {
+        __pyx_t_4 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_1 = PyList_GET_ITEM(sequence, 1); 
+      }
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(__pyx_t_1);
+      #else
+      __pyx_t_4 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 583, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_1 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 583, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      #endif
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      __pyx_t_7 = PyObject_GetIter(__pyx_t_10); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 583, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      __pyx_t_12 = Py_TYPE(__pyx_t_7)->tp_iternext;
+      index = 0; __pyx_t_4 = __pyx_t_12(__pyx_t_7); if (unlikely(!__pyx_t_4)) goto __pyx_L12_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_4);
+      index = 1; __pyx_t_1 = __pyx_t_12(__pyx_t_7); if (unlikely(!__pyx_t_1)) goto __pyx_L12_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_1);
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_7), 2) < 0) __PYX_ERR(0, 583, __pyx_L1_error)
+      __pyx_t_12 = NULL;
+      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      goto __pyx_L13_unpacking_done;
+      __pyx_L12_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      __pyx_t_12 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 583, __pyx_L1_error)
+      __pyx_L13_unpacking_done:;
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_4);
+    __pyx_t_4 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_summary, __pyx_t_1);
+    __pyx_t_1 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0584:             tree.summary['group_size'] += ' ' + treatment_group + ': ' + str(summary[1])
+
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __Pyx_INCREF(__pyx_n_s_group_size);
+    __pyx_t_15 = __pyx_n_s_group_size;
+    __pyx_t_1 = __Pyx_PyObject_Dict_GetItem(__pyx_t_10, __pyx_t_15); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = PyNumber_Add(__pyx_kp_s__5, __pyx_v_treatment_group); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_7 = PyNumber_Add(__pyx_t_4, __pyx_kp_s__6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = __Pyx_GetItemInt(__pyx_v_summary, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyString_Type)), __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = PyNumber_Add(__pyx_t_7, __pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_1, __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (unlikely(PyObject_SetItem(__pyx_t_10, __pyx_t_15, __pyx_t_9) < 0)) __PYX_ERR(0, 584, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
 0585:         # classProb
+
+0586:         if tree.results is not None:
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 586, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = (__pyx_t_2 != Py_None);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_6 = (__pyx_t_5 != 0);
+  if (__pyx_t_6) {
+/* … */
+  }
+
+0587:             tree.results = self.uplift_classification_results(treatment_idx, y)
+
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 587, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __pyx_t_9 = NULL;
+    __pyx_t_8 = 0;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+      __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_10);
+      if (likely(__pyx_t_9)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+        __Pyx_INCREF(__pyx_t_9);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_10, function);
+        __pyx_t_8 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_10)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_v_treatment_idx, __pyx_v_y};
+      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __Pyx_GOTREF(__pyx_t_2);
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_v_treatment_idx, __pyx_v_y};
+      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __Pyx_GOTREF(__pyx_t_2);
+    } else
+    #endif
+    {
+      __pyx_t_4 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 587, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      if (__pyx_t_9) {
+        __Pyx_GIVEREF(__pyx_t_9); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_9); __pyx_t_9 = NULL;
+      }
+      __Pyx_INCREF(__pyx_v_treatment_idx);
+      __Pyx_GIVEREF(__pyx_v_treatment_idx);
+      PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_8, __pyx_v_treatment_idx);
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_8, __pyx_v_y);
+      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_4, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_2) < 0) __PYX_ERR(0, 587, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0588:         return self
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_self);
+  __pyx_r = __pyx_v_self;
+  goto __pyx_L0;
+
 0589:
+
+0590:     def predict(self, X):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict[] = "\n        Returns the recommended treatment group and predicted optimal\n        probability conditional on using the recommended treatment group.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        Returns\n        -------\n        pred: ndarray, shape = [num_samples, num_treatments]\n            An ndarray of predicted treatment effects across treatments.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,0};
+    PyObject* values[2] = {0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, 1); __PYX_ERR(0, 590, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 590, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 590, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict(__pyx_self, __pyx_v_self, __pyx_v_X);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X) {
+  PyObject *__pyx_v_pred_nodes = NULL;
+  Py_ssize_t __pyx_v_i_row;
+  PyObject *__pyx_v_pred_leaf = NULL;
+  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict", 0);
+  __Pyx_INCREF(__pyx_v_X);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_pred_nodes);
+  __Pyx_XDECREF(__pyx_v_pred_leaf);
+  __Pyx_XDECREF(__pyx_v__);
+  __Pyx_XDECREF(__pyx_v_X);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__35 = PyTuple_Pack(6, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_pred_nodes, __pyx_n_s_i_row, __pyx_n_s_pred_leaf, __pyx_n_s__34); if (unlikely(!__pyx_tuple__35)) __PYX_ERR(0, 590, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__35);
+  __Pyx_GIVEREF(__pyx_tuple__35);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict, 0, __pyx_n_s_UpliftTreeClassifier_predict, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__36)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 590, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_predict, __pyx_t_2) < 0) __PYX_ERR(0, 590, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_codeobj__36 = (PyObject*)__Pyx_PyCode_New(2, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__35, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_predict, 590, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__36)) __PYX_ERR(0, 590, __pyx_L1_error)
+
 0591:         '''
+
 0592:         Returns the recommended treatment group and predicted optimal
+
 0593:         probability conditional on using the recommended treatment group.
+
 0594:
+
 0595:         Args
+
 0596:         ----
+
 0597:         X : ndarray, shape = [num_samples, num_features]
+
 0598:             An ndarray of the covariates used to train the uplift model.
+
 0599:
+
 0600:         Returns
+
 0601:         -------
+
 0602:         pred: ndarray, shape = [num_samples, num_treatments]
+
 0603:             An ndarray of predicted treatment effects across treatments.
+
 0604:         '''
+
 0605:
+
+0606:         X = check_array(X)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_array); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 606, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 606, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_1);
+  __pyx_t_1 = 0;
+
 0607:
+
+0608:         pred_nodes = []
+
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 608, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_pred_nodes = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
+0609:         for i_row in range(len(X)):
+
  __pyx_t_4 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_4 == ((Py_ssize_t)-1))) __PYX_ERR(0, 609, __pyx_L1_error)
+  __pyx_t_5 = __pyx_t_4;
+  for (__pyx_t_6 = 0; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
+    __pyx_v_i_row = __pyx_t_6;
+
+0610:             pred_leaf, _ = self.classify(X[i_row], self.fitted_uplift_tree, dataMissing=False)
+
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classify); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_X, __pyx_v_i_row, Py_ssize_t, 1, PyInt_FromSsize_t, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_7 = PyTuple_New(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __Pyx_GIVEREF(__pyx_t_2);
+    PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_2);
+    __Pyx_GIVEREF(__pyx_t_3);
+    PyTuple_SET_ITEM(__pyx_t_7, 1, __pyx_t_3);
+    __pyx_t_2 = 0;
+    __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_dataMissing, Py_False) < 0) __PYX_ERR(0, 610, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_7, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 610, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if ((likely(PyTuple_CheckExact(__pyx_t_2))) || (PyList_CheckExact(__pyx_t_2))) {
+      PyObject* sequence = __pyx_t_2;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 2)) {
+        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 610, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_7 = PyTuple_GET_ITEM(sequence, 1); 
+      } else {
+        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_7 = PyList_GET_ITEM(sequence, 1); 
+      }
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_7);
+      #else
+      __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_7 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 610, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      #endif
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 610, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __pyx_t_8 = Py_TYPE(__pyx_t_1)->tp_iternext;
+      index = 0; __pyx_t_3 = __pyx_t_8(__pyx_t_1); if (unlikely(!__pyx_t_3)) goto __pyx_L5_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_3);
+      index = 1; __pyx_t_7 = __pyx_t_8(__pyx_t_1); if (unlikely(!__pyx_t_7)) goto __pyx_L5_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_7);
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_8(__pyx_t_1), 2) < 0) __PYX_ERR(0, 610, __pyx_L1_error)
+      __pyx_t_8 = NULL;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      goto __pyx_L6_unpacking_done;
+      __pyx_L5_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_t_8 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 610, __pyx_L1_error)
+      __pyx_L6_unpacking_done:;
+    }
+    __Pyx_XDECREF_SET(__pyx_v_pred_leaf, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_7);
+    __pyx_t_7 = 0;
+
+0611:             pred_nodes.append(pred_leaf)
+
    __pyx_t_9 = __Pyx_PyList_Append(__pyx_v_pred_nodes, __pyx_v_pred_leaf); if (unlikely(__pyx_t_9 == ((int)-1))) __PYX_ERR(0, 611, __pyx_L1_error)
+  }
+
+0612:         return np.array(pred_nodes)
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_np); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 612, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_array); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 612, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_t_7 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_7)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_7);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_7, __pyx_v_pred_nodes) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_pred_nodes);
+  __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 612, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_r = __pyx_t_2;
+  __pyx_t_2 = 0;
+  goto __pyx_L0;
+
 0613:
+
+0614:     @staticmethod
+
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 614, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_divideSet, __pyx_t_3) < 0) __PYX_ERR(0, 615, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0615:     def divideSet(X, treatment_idx, y, column, value):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet[] = "\n        Tree node split.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        column : int\n                The column used to split the data.\n        value : float or int\n                The value in the column for splitting the data.\n\n        Returns\n        -------\n        (X_l, X_r, treatment_l, treatment_r, y_l, y_r) : list of ndarray\n                The covariates, treatments and outcomes of left node and the right node.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet = {"divideSet", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_column = 0;
+  PyObject *__pyx_v_value = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("divideSet (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_column,&__pyx_n_s_value,0};
+    PyObject* values[5] = {0,0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 1); __PYX_ERR(0, 615, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 2); __PYX_ERR(0, 615, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_column)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 3); __PYX_ERR(0, 615, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_value)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 4); __PYX_ERR(0, 615, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "divideSet") < 0)) __PYX_ERR(0, 615, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 5) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+      values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+    }
+    __pyx_v_X = values[0];
+    __pyx_v_treatment_idx = values[1];
+    __pyx_v_y = values[2];
+    __pyx_v_column = values[3];
+    __pyx_v_value = values[4];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 615, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.divideSet", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet(__pyx_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_column, __pyx_v_value);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_column, PyObject *__pyx_v_value) {
+  PyObject *__pyx_v_filt = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("divideSet", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.divideSet", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_filt);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__37 = PyTuple_Pack(6, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_column, __pyx_n_s_value, __pyx_n_s_filt); if (unlikely(!__pyx_tuple__37)) __PYX_ERR(0, 615, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__37);
+  __Pyx_GIVEREF(__pyx_tuple__37);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_divideSet, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__38)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 615, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_codeobj__38 = (PyObject*)__Pyx_PyCode_New(5, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__37, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_divideSet, 615, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__38)) __PYX_ERR(0, 615, __pyx_L1_error)
+
 0616:         '''
+
 0617:         Tree node split.
+
 0618:
+
 0619:         Args
+
 0620:         ----
+
 0621:         X : ndarray, shape = [num_samples, num_features]
+
 0622:             An ndarray of the covariates used to train the uplift model.
+
 0623:         treatment_idx : array-like, shape = [num_samples]
+
 0624:             An array containing the treatment group index for each unit.
+
 0625:         y : array-like, shape = [num_samples]
+
 0626:             An array containing the outcome of interest for each unit.
+
 0627:         column : int
+
 0628:                 The column used to split the data.
+
 0629:         value : float or int
+
 0630:                 The value in the column for splitting the data.
+
 0631:
+
 0632:         Returns
+
 0633:         -------
+
 0634:         (X_l, X_r, treatment_l, treatment_r, y_l, y_r) : list of ndarray
+
 0635:                 The covariates, treatments and outcomes of left node and the right node.
+
 0636:         '''
+
 0637:         # for int and float values
+
+0638:         if np.issubdtype(value.dtype, np.number):
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 638, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_issubdtype); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 638, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_dtype); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 638, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = NULL;
+  __pyx_t_5 = 0;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+      __pyx_t_5 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_3)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_2, ((PyObject *)__pyx_ptype_5numpy_number)};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_2, ((PyObject *)__pyx_ptype_5numpy_number)};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  } else
+  #endif
+  {
+    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 638, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    if (__pyx_t_4) {
+      __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_4); __pyx_t_4 = NULL;
+    }
+    __Pyx_GIVEREF(__pyx_t_2);
+    PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_5, __pyx_t_2);
+    __Pyx_INCREF(((PyObject *)__pyx_ptype_5numpy_number));
+    __Pyx_GIVEREF(((PyObject *)__pyx_ptype_5numpy_number));
+    PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, ((PyObject *)__pyx_ptype_5numpy_number));
+    __pyx_t_2 = 0;
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 638, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (__pyx_t_7) {
+/* … */
+    goto __pyx_L3;
+  }
+
+0639:             filt = X[:, column] >= value
+
    __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 639, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_INCREF(__pyx_slice__7);
+    __Pyx_GIVEREF(__pyx_slice__7);
+    PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_slice__7);
+    __Pyx_INCREF(__pyx_v_column);
+    __Pyx_GIVEREF(__pyx_v_column);
+    PyTuple_SET_ITEM(__pyx_t_1, 1, __pyx_v_column);
+    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 639, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = PyObject_RichCompare(__pyx_t_3, __pyx_v_value, Py_GE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 639, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_v_filt = __pyx_t_1;
+    __pyx_t_1 = 0;
+/* … */
+  __pyx_slice__7 = PySlice_New(Py_None, Py_None, Py_None); if (unlikely(!__pyx_slice__7)) __PYX_ERR(0, 639, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_slice__7);
+  __Pyx_GIVEREF(__pyx_slice__7);
+
 0640:         else:  # for strings
+
+0641:             filt = X[:, column] == value
+
  /*else*/ {
+    __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 641, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_INCREF(__pyx_slice__7);
+    __Pyx_GIVEREF(__pyx_slice__7);
+    PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_slice__7);
+    __Pyx_INCREF(__pyx_v_column);
+    __Pyx_GIVEREF(__pyx_v_column);
+    PyTuple_SET_ITEM(__pyx_t_1, 1, __pyx_v_column);
+    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 641, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = PyObject_RichCompare(__pyx_t_3, __pyx_v_value, Py_EQ); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 641, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_v_filt = __pyx_t_1;
+    __pyx_t_1 = 0;
+  }
+  __pyx_L3:;
+
 0642:
+
+0643:         return X[filt], X[~filt], treatment_idx[filt], treatment_idx[~filt], y[filt], y[~filt]
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_filt); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_6 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_treatment_idx, __pyx_v_filt); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_2 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_treatment_idx, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_filt); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_8 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __pyx_t_9 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  __pyx_t_8 = PyTuple_New(6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 643, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __Pyx_GIVEREF(__pyx_t_1);
+  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_t_1);
+  __Pyx_GIVEREF(__pyx_t_6);
+  PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_t_6);
+  __Pyx_GIVEREF(__pyx_t_3);
+  PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_t_3);
+  __Pyx_GIVEREF(__pyx_t_4);
+  PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_4);
+  __Pyx_GIVEREF(__pyx_t_2);
+  PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_t_2);
+  __Pyx_GIVEREF(__pyx_t_9);
+  PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_t_9);
+  __pyx_t_1 = 0;
+  __pyx_t_6 = 0;
+  __pyx_t_3 = 0;
+  __pyx_t_4 = 0;
+  __pyx_t_2 = 0;
+  __pyx_t_9 = 0;
+  __pyx_r = __pyx_t_8;
+  __pyx_t_8 = 0;
+  goto __pyx_L0;
+
 0644:
+
+0645:     def group_uniqueCounts(self, treatment_idx, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts[] = "\n        Count sample size by experiment group.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        results : list of list\n            The negative and positive outcome sample sizes for each of the control and treatment groups.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts = {"group_uniqueCounts", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("group_uniqueCounts (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,0};
+    PyObject* values[3] = {0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, 1); __PYX_ERR(0, 645, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, 2); __PYX_ERR(0, 645, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "group_uniqueCounts") < 0)) __PYX_ERR(0, 645, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_treatment_idx = values[1];
+    __pyx_v_y = values[2];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 645, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.group_uniqueCounts", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_results = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_filt = NULL;
+  PyObject *__pyx_v_n_pos = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("group_uniqueCounts", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.group_uniqueCounts", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_results);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_filt);
+  __Pyx_XDECREF(__pyx_v_n_pos);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__39 = PyTuple_Pack(7, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_results, __pyx_n_s_i, __pyx_n_s_filt, __pyx_n_s_n_pos); if (unlikely(!__pyx_tuple__39)) __PYX_ERR(0, 645, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__39);
+  __Pyx_GIVEREF(__pyx_tuple__39);
+/* … */
+  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts, 0, __pyx_n_s_UpliftTreeClassifier_group_uniqu, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__40)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 645, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_group_uniqueCounts, __pyx_t_3) < 0) __PYX_ERR(0, 645, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_codeobj__40 = (PyObject*)__Pyx_PyCode_New(3, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__39, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_group_uniqueCounts, 645, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__40)) __PYX_ERR(0, 645, __pyx_L1_error)
+
 0646:         '''
+
 0647:         Count sample size by experiment group.
+
 0648:
+
 0649:         Args
+
 0650:         ----
+
 0651:         treatment_idx : array-like, shape = [num_samples]
+
 0652:             An array containing the treatment group index for each unit.
+
 0653:         y : array-like, shape = [num_samples]
+
 0654:             An array containing the outcome of interest for each unit.
+
 0655:
+
 0656:         Returns
+
 0657:         -------
+
 0658:         results : list of list
+
 0659:             The negative and positive outcome sample sizes for each of the control and treatment groups.
+
 0660:         '''
+
+0661:         results = []
+
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 661, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_results = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
+0662:         for i in range(self.n_class):
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 662, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 662, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 662, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 662, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 662, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 662, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0663:             filt = treatment_idx == i
+
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment_idx, __pyx_v_i, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 663, __pyx_L1_error)
+    __Pyx_XDECREF_SET(__pyx_v_filt, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0664:             n_pos = y[filt].sum()
+
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_filt); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 664, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_sum); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 664, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = NULL;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
+      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
+      if (likely(__pyx_t_5)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+        __Pyx_INCREF(__pyx_t_5);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_6, function);
+      }
+    }
+    __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_6);
+    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 664, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_2);
+    __pyx_t_2 = 0;
+
 0665:
+
 0666:             # [N(Y = 0, T = 1), N(Y = 1, T = 1)]
+
+0667:             results.append([filt.sum() - n_pos, n_pos])
+
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_filt, __pyx_n_s_sum); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 667, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_5 = NULL;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
+      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
+      if (likely(__pyx_t_5)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+        __Pyx_INCREF(__pyx_t_5);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_6, function);
+      }
+    }
+    __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_6);
+    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 667, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_6 = PyNumber_Subtract(__pyx_t_2, __pyx_v_n_pos); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 667, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 667, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_GIVEREF(__pyx_t_6);
+    PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_6);
+    __Pyx_INCREF(__pyx_v_n_pos);
+    __Pyx_GIVEREF(__pyx_v_n_pos);
+    PyList_SET_ITEM(__pyx_t_2, 1, __pyx_v_n_pos);
+    __pyx_t_6 = 0;
+    __pyx_t_7 = __Pyx_PyList_Append(__pyx_v_results, __pyx_t_2); if (unlikely(__pyx_t_7 == ((int)-1))) __PYX_ERR(0, 667, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 0668:
+
+0669:         return results
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_results);
+  __pyx_r = __pyx_v_results;
+  goto __pyx_L0;
+
 0670:
+
+0671:     @staticmethod
+
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 671, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_KL, __pyx_t_2) < 0) __PYX_ERR(0, 672, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0672:     def evaluate_KL(nodeSummary):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL[] = "\n        Calculate KL Divergence as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary()\n            method.\n\n        Returns\n        -------\n        d_res : KL Divergence\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL = {"evaluate_KL", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_KL (wrapper)", 0);
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_v_p_c = NULL;
+  double __pyx_v_d_res;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_KL", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_KL", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_p_c);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__41 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_p_c, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__41)) __PYX_ERR(0, 672, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__41);
+  __Pyx_GIVEREF(__pyx_tuple__41);
+/* … */
+  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_KL, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__42)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 672, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_codeobj__42 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__41, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_KL, 672, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__42)) __PYX_ERR(0, 672, __pyx_L1_error)
+
 0673:         '''
+
 0674:         Calculate KL Divergence as split evaluation criterion for a given node.
+
 0675:
+
 0676:         Args
+
 0677:         ----
+
 0678:         nodeSummary : list of list
+
 0679:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary()
+
 0680:             method.
+
 0681:
+
 0682:         Returns
+
 0683:         -------
+
 0684:         d_res : KL Divergence
+
 0685:         '''
+
+0686:         p_c = nodeSummary[0][0]
+
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 686, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 686, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_p_c = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+0687:         d_res = 0.
+
  __pyx_v_d_res = 0.;
+
+0688:         for treatment_group in nodeSummary[1:]:
+
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 688, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 688, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 688, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 688, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 688, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0689:             d_res += kl_divergence(treatment_group[0], p_c)
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 689, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_2); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_v_p_c); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
+    __pyx_t_7 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_t_5, __pyx_t_6); if (unlikely(__pyx_t_7 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
+    __pyx_v_d_res = (__pyx_v_d_res + __pyx_t_7);
+
+0690:         return d_res
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_1 = PyFloat_FromDouble(__pyx_v_d_res); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 690, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_r = __pyx_t_1;
+  __pyx_t_1 = 0;
+  goto __pyx_L0;
+
 0691:
+
+0692:     @staticmethod
+
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 692, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_ED, __pyx_t_3) < 0) __PYX_ERR(0, 693, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0693:     def evaluate_ED(nodeSummary):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED[] = "\n        Calculate Euclidean Distance as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : dictionary\n            The tree node summary statistics, produced by tree_node_summary()\n            method.\n\n        Returns\n        -------\n        d_res : Euclidean Distance\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED = {"evaluate_ED", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_ED (wrapper)", 0);
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_v_pc = NULL;
+  PyObject *__pyx_v_d_res = NULL;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_ED", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_ED", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_pc);
+  __Pyx_XDECREF(__pyx_v_d_res);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__43 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__43)) __PYX_ERR(0, 693, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__43);
+  __Pyx_GIVEREF(__pyx_tuple__43);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_ED, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__44)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 693, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_codeobj__44 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__43, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_ED, 693, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__44)) __PYX_ERR(0, 693, __pyx_L1_error)
+
 0694:         '''
+
 0695:         Calculate Euclidean Distance as split evaluation criterion for a given node.
+
 0696:
+
 0697:         Args
+
 0698:         ----
+
 0699:         nodeSummary : dictionary
+
 0700:             The tree node summary statistics, produced by tree_node_summary()
+
 0701:             method.
+
 0702:
+
 0703:         Returns
+
 0704:         -------
+
 0705:         d_res : Euclidean Distance
+
 0706:         '''
+
+0707:         pc = nodeSummary[0][0]
+
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 707, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 707, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_pc = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+0708:         d_res = 0
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_d_res = __pyx_int_0;
+
+0709:         for treatment_group in nodeSummary[1:]:
+
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 709, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 709, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 709, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 709, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 709, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0710:             d_res += 2*(treatment_group[0] - pc)**2
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 710, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyNumber_Power(__pyx_t_5, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = PyNumber_Multiply(__pyx_int_2, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 710, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0711:         return d_res
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_d_res);
+  __pyx_r = __pyx_v_d_res;
+  goto __pyx_L0;
+
 0712:
+
+0713:     @staticmethod
+
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 713, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_Chi, __pyx_t_2) < 0) __PYX_ERR(0, 714, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0714:     def evaluate_Chi(nodeSummary):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi[] = "\n        Calculate Chi-Square statistic as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : dictionary\n            The tree node summary statistics, produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Chi-Square\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi = {"evaluate_Chi", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_Chi (wrapper)", 0);
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_v_pc = NULL;
+  PyObject *__pyx_v_d_res = NULL;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_Chi", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_Chi", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_pc);
+  __Pyx_XDECREF(__pyx_v_d_res);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__45 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__45)) __PYX_ERR(0, 714, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__45);
+  __Pyx_GIVEREF(__pyx_tuple__45);
+/* … */
+  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_Ch, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__46)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 714, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_codeobj__46 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__45, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_Chi, 714, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__46)) __PYX_ERR(0, 714, __pyx_L1_error)
+
 0715:         '''
+
 0716:         Calculate Chi-Square statistic as split evaluation criterion for a given node.
+
 0717:
+
 0718:         Args
+
 0719:         ----
+
 0720:         nodeSummary : dictionary
+
 0721:             The tree node summary statistics, produced by tree_node_summary() method.
+
 0722:
+
 0723:         Returns
+
 0724:         -------
+
 0725:         d_res : Chi-Square
+
 0726:         '''
+
+0727:         pc = nodeSummary[0][0]
+
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 727, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 727, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_pc = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+0728:         d_res = 0
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_d_res = __pyx_int_0;
+
+0729:         for treatment_group in nodeSummary[1:]:
+
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 729, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 729, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 729, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 729, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 729, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0730:             d_res += ((treatment_group[0] - pc) ** 2 / max(0.1 ** 6, pc)
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyNumber_Power(__pyx_t_5, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_INCREF(__pyx_v_pc);
+    __pyx_t_5 = __pyx_v_pc;
+    __pyx_t_6 = pow(0.1, 6.0);
+    __pyx_t_8 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __pyx_t_9 = PyObject_RichCompare(__pyx_t_5, __pyx_t_8, Py_GT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    if (__pyx_t_10) {
+      __Pyx_INCREF(__pyx_t_5);
+      __pyx_t_7 = __pyx_t_5;
+    } else {
+      __pyx_t_9 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 730, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __pyx_t_7 = __pyx_t_9;
+      __pyx_t_9 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_7); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+/* … */
+    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_9); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0731:                       + (treatment_group[0] - pc) ** 2 / max(0.1 ** 6, 1 - pc))
+
    __pyx_t_7 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __pyx_t_2 = PyNumber_Subtract(__pyx_t_7, __pyx_v_pc); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __pyx_t_7 = PyNumber_Power(__pyx_t_2, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_pc, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_6 = pow(0.1, 6.0);
+    __pyx_t_8 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __pyx_t_11 = PyObject_RichCompare(__pyx_t_2, __pyx_t_8, Py_GT); __Pyx_XGOTREF(__pyx_t_11); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_11); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+    if (__pyx_t_10) {
+      __Pyx_INCREF(__pyx_t_2);
+      __pyx_t_9 = __pyx_t_2;
+    } else {
+      __pyx_t_11 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 731, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_9 = __pyx_t_11;
+      __pyx_t_11 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_7, __pyx_t_9); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_t_9 = PyNumber_Add(__pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 731, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0732:         return d_res
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_d_res);
+  __pyx_r = __pyx_v_d_res;
+  goto __pyx_L0;
+
 0733:
+
+0734:     @staticmethod
+
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 734, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_DDP, __pyx_t_3) < 0) __PYX_ERR(0, 735, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+0735:     def evaluate_DDP(nodeSummary):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP[] = "\n        Calculate Delta P as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Delta P\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP = {"evaluate_DDP", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_DDP (wrapper)", 0);
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_v_pc = NULL;
+  PyObject *__pyx_v_d_res = NULL;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_DDP", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_DDP", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_pc);
+  __Pyx_XDECREF(__pyx_v_d_res);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__47 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__47)) __PYX_ERR(0, 735, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__47);
+  __Pyx_GIVEREF(__pyx_tuple__47);
+/* … */
+  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_DD, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__48)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 735, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_codeobj__48 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__47, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_DDP, 735, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__48)) __PYX_ERR(0, 735, __pyx_L1_error)
+
 0736:         '''
+
 0737:         Calculate Delta P as split evaluation criterion for a given node.
+
 0738:
+
 0739:         Args
+
 0740:         ----
+
 0741:         nodeSummary : list of list
+
 0742:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.
+
 0743:
+
 0744:         Returns
+
 0745:         -------
+
 0746:         d_res : Delta P
+
 0747:         '''
+
+0748:         pc = nodeSummary[0][0]
+
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 748, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 748, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_pc = __pyx_t_2;
+  __pyx_t_2 = 0;
+
+0749:         d_res = 0
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_d_res = __pyx_int_0;
+
+0750:         for treatment_group in nodeSummary[1:]:
+
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 750, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 750, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 750, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 750, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 750, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0751:             d_res += treatment_group[0] - pc
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 751, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0752:         return d_res
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_d_res);
+  __pyx_r = __pyx_v_d_res;
+  goto __pyx_L0;
+
 0753:
+
+0754:     @staticmethod
+
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 754, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_CTS, __pyx_t_2) < 0) __PYX_ERR(0, 755, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+0755:     def evaluate_CTS(nodeSummary):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS[] = "\n        Calculate CTS (conditional treatment selection) as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Chi-Square\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS = {"evaluate_CTS", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_CTS (wrapper)", 0);
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
+  PyObject *__pyx_v_stat = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("evaluate_CTS", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_CTS", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_stat);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__49 = PyTuple_Pack(2, __pyx_n_s_nodeSummary, __pyx_n_s_stat); if (unlikely(!__pyx_tuple__49)) __PYX_ERR(0, 755, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__49);
+  __Pyx_GIVEREF(__pyx_tuple__49);
+/* … */
+  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_CT, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__50)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 755, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_codeobj__50 = (PyObject*)__Pyx_PyCode_New(1, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__49, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_CTS, 755, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__50)) __PYX_ERR(0, 755, __pyx_L1_error)
+
 0756:         '''
+
 0757:         Calculate CTS (conditional treatment selection) as split evaluation criterion for a given node.
+
 0758:
+
 0759:         Args
+
 0760:         ----
+
 0761:         nodeSummary : list of list
+
 0762:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.
+
 0763:
+
 0764:         Returns
+
 0765:         -------
+
 0766:         d_res : Chi-Square
+
 0767:         '''
+
+0768:         return -max([stat[0] for stat in nodeSummary])
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 768, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (likely(PyList_CheckExact(__pyx_v_nodeSummary)) || PyTuple_CheckExact(__pyx_v_nodeSummary)) {
+    __pyx_t_2 = __pyx_v_nodeSummary; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_nodeSummary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 768, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 768, __pyx_L1_error)
+  }
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 768, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 768, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      }
+    } else {
+      __pyx_t_5 = __pyx_t_4(__pyx_t_2);
+      if (unlikely(!__pyx_t_5)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 768, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_5);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_stat, __pyx_t_5);
+    __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_stat, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 768, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_max, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 768, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = PyNumber_Negative(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 768, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_r = __pyx_t_1;
+  __pyx_t_1 = 0;
+  goto __pyx_L0;
+
 0769:
+
+0770:     def normI(self, n_c: cython.int, n_c_left: cython.int, n_t: list, n_t_left: list,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI[] = "\n        Normalization factor.\n\n        Args\n        ----\n        currentNodeSummary : list of list\n            The summary statistics of the current tree node, [P(Y=1|T), N(T)].\n\n        leftNodeSummary : list of list\n            The summary statistics of the left tree node, [P(Y=1|T), N(T)].\n\n        alpha : float\n            The weight used to balance different normalization parts.\n\n        Returns\n        -------\n        norm_res : float\n            Normalization factor.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI = {"normI", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  CYTHON_UNUSED PyObject *__pyx_v_self = 0;
+  int __pyx_v_n_c;
+  int __pyx_v_n_c_left;
+  PyObject *__pyx_v_n_t = 0;
+  PyObject *__pyx_v_n_t_left = 0;
+  float __pyx_v_alpha;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("normI (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_n_c,&__pyx_n_s_n_c_left,&__pyx_n_s_n_t,&__pyx_n_s_n_t_left,&__pyx_n_s_alpha,0};
+    PyObject* values[6] = {0,0,0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_c)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 1); __PYX_ERR(0, 770, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_c_left)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 2); __PYX_ERR(0, 770, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_t)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 3); __PYX_ERR(0, 770, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_t_left)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 4); __PYX_ERR(0, 770, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_alpha);
+          if (value) { values[5] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "normI") < 0)) __PYX_ERR(0, 770, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_n_c = __Pyx_PyInt_As_int(values[1]); if (unlikely((__pyx_v_n_c == (int)-1) && PyErr_Occurred())) __PYX_ERR(0, 770, __pyx_L3_error)
+    __pyx_v_n_c_left = __Pyx_PyInt_As_int(values[2]); if (unlikely((__pyx_v_n_c_left == (int)-1) && PyErr_Occurred())) __PYX_ERR(0, 770, __pyx_L3_error)
+    __pyx_v_n_t = ((PyObject*)values[3]);
+    __pyx_v_n_t_left = ((PyObject*)values[4]);
+    if (values[5]) {
+      __pyx_v_alpha = __pyx_PyFloat_AsFloat(values[5]); if (unlikely((__pyx_v_alpha == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 771, __pyx_L3_error)
+    } else {
+      __pyx_v_alpha = ((float)((double)0.9));
+    }
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 770, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.normI", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_n_t), (&PyList_Type), 1, "n_t", 1))) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_n_t_left), (&PyList_Type), 1, "n_t_left", 1))) __PYX_ERR(0, 770, __pyx_L1_error)
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI(__pyx_self, __pyx_v_self, __pyx_v_n_c, __pyx_v_n_c_left, __pyx_v_n_t, __pyx_v_n_t_left, __pyx_v_alpha);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI(CYTHON_UNUSED PyObject *__pyx_self, CYTHON_UNUSED PyObject *__pyx_v_self, int __pyx_v_n_c, int __pyx_v_n_c_left, PyObject *__pyx_v_n_t, PyObject *__pyx_v_n_t_left, float __pyx_v_alpha) {
+  float __pyx_v_norm_res;
+  float __pyx_v_pt_a;
+  float __pyx_v_pc_a;
+  Py_ssize_t __pyx_v_i;
+  PyObject *__pyx_v_pt_a_i = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("normI", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_12);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.normI", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_pt_a_i);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__51 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_n_c, __pyx_n_s_n_c_left, __pyx_n_s_n_t, __pyx_n_s_n_t_left, __pyx_n_s_alpha, __pyx_n_s_norm_res, __pyx_n_s_pt_a, __pyx_n_s_pc_a, __pyx_n_s_i, __pyx_n_s_pt_a_i); if (unlikely(!__pyx_tuple__51)) __PYX_ERR(0, 770, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__51);
+  __Pyx_GIVEREF(__pyx_tuple__51);
+/* … */
+  __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 770, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_GIVEREF(__pyx_t_2);
+  PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
+  __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 770, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_c, __pyx_n_u_int) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_c_left, __pyx_n_u_int) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_t, __pyx_n_u_list) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_t_left, __pyx_n_u_list) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_alpha, __pyx_n_u_float) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_return, __pyx_n_u_float) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI, 0, __pyx_n_s_UpliftTreeClassifier_normI, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__52)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 770, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_t_3);
+  __Pyx_CyFunction_SetAnnotationsDict(__pyx_t_5, __pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_normI, __pyx_t_5) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_codeobj__52 = (PyObject*)__Pyx_PyCode_New(6, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__51, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_normI, 770, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__52)) __PYX_ERR(0, 770, __pyx_L1_error)
+
+0771:               alpha: cython.float=0.9) -> cython.float:
+
  __pyx_t_2 = PyFloat_FromDouble(((double)0.9)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 771, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+
 0772:         '''
+
 0773:         Normalization factor.
+
 0774:
+
 0775:         Args
+
 0776:         ----
+
 0777:         currentNodeSummary : list of list
+
 0778:             The summary statistics of the current tree node, [P(Y=1|T), N(T)].
+
 0779:
+
 0780:         leftNodeSummary : list of list
+
 0781:             The summary statistics of the left tree node, [P(Y=1|T), N(T)].
+
 0782:
+
 0783:         alpha : float
+
 0784:             The weight used to balance different normalization parts.
+
 0785:
+
 0786:         Returns
+
 0787:         -------
+
 0788:         norm_res : float
+
 0789:             Normalization factor.
+
 0790:         '''
+
 0791:
+
+0792:         norm_res: cython.float = 0.
+
  __pyx_v_norm_res = 0.;
+
 0793:         pt_a: cython.float
+
 0794:         pc_a: cython.float
+
 0795:
+
+0796:         pt_a = 1. * np.sum(n_t_left) / (np.sum(n_t) + 0.1)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_2, __pyx_v_n_t_left) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t_left);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = PyNumber_Multiply(__pyx_float_1_, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_4, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_2, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = __Pyx_PyFloat_AddObjC(__pyx_t_1, __pyx_float_0_1, 0.1, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 796, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_v_pt_a = __pyx_t_5;
+
+0797:         pc_a = 1. * n_c_left / (n_c + 0.1)
+
  __pyx_v_pc_a = ((1. * __pyx_v_n_c_left) / (__pyx_v_n_c + 0.1));
+
 0798:
+
 0799:         # Normalization Part 1
+
+0800:         norm_res += (
+
  __pyx_v_norm_res = (__pyx_v_norm_res + ((__pyx_v_alpha * __pyx_t_7) * __pyx_t_6));
+
+0801:             alpha * entropyH(1. * np.sum(n_t) / (np.sum(n_t) + n_c), 1. * n_c / (np.sum(n_t) + n_c))
+
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = PyNumber_Multiply(__pyx_float_1_, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_4 = PyNumber_Add(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_2); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_sum); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_1, function);
+    }
+  }
+  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_3 = PyNumber_Add(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_8.__pyx_n = 1;
+  __pyx_t_8.q = __pyx_t_6;
+  __pyx_t_7 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_5, &__pyx_t_8); if (unlikely(__pyx_t_7 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
+
+0802:             * kl_divergence(pt_a, pc_a)
+
  __pyx_t_6 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_v_pt_a, __pyx_v_pc_a); if (unlikely(__pyx_t_6 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 802, __pyx_L1_error)
+
 0803:         )
+
 0804:         # Normalization Part 2 & 3
+
+0805:         for i in range(len(n_t)):
+
  if (unlikely(__pyx_v_n_t == Py_None)) {
+    PyErr_SetString(PyExc_TypeError, "object of type 'NoneType' has no len()");
+    __PYX_ERR(0, 805, __pyx_L1_error)
+  }
+  __pyx_t_9 = PyList_GET_SIZE(__pyx_v_n_t); if (unlikely(__pyx_t_9 == ((Py_ssize_t)-1))) __PYX_ERR(0, 805, __pyx_L1_error)
+  __pyx_t_10 = __pyx_t_9;
+  for (__pyx_t_11 = 0; __pyx_t_11 < __pyx_t_10; __pyx_t_11+=1) {
+    __pyx_v_i = __pyx_t_11;
+
+0806:             pt_a_i = 1. * n_t_left[i] / (n_t[i] + 0.1)
+
    if (unlikely(__pyx_v_n_t_left == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 806, __pyx_L1_error)
+    }
+    __pyx_t_1 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t_left, __pyx_v_i)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 806, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    if (unlikely(__pyx_v_n_t == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 806, __pyx_L1_error)
+    }
+    __pyx_t_3 = __Pyx_PyFloat_AddObjC(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_float_0_1, 0.1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 806, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 806, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_pt_a_i, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0807:             norm_res += (
+
    __pyx_v_norm_res = (__pyx_v_norm_res + (((1.0 - __pyx_v_alpha) * __pyx_t_5) * __pyx_t_6));
+
+0808:                 (1 - alpha) * entropyH(1. * n_t[i] / (n_t[i] + n_c), 1. * n_c / (n_t[i] + n_c))
+
    if (unlikely(__pyx_v_n_t == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 808, __pyx_L1_error)
+    }
+    __pyx_t_2 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (unlikely(__pyx_v_n_t == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 808, __pyx_L1_error)
+    }
+    __pyx_t_3 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_1 = PyNumber_Add(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_t_3); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    if (unlikely(__pyx_v_n_t == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 808, __pyx_L1_error)
+    }
+    __pyx_t_1 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_2 = PyNumber_Add(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_7 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_7 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_8.__pyx_n = 1;
+    __pyx_t_8.q = __pyx_t_7;
+    __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_6, &__pyx_t_8); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
+
+0809:                 * kl_divergence(1. * pt_a_i, pc_a)
+
    __pyx_t_1 = PyNumber_Multiply(__pyx_float_1_, __pyx_v_pt_a_i); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 809, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_7 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_7 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 809, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_t_6 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_t_7, __pyx_v_pc_a); if (unlikely(__pyx_t_6 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 809, __pyx_L1_error)
+
 0810:             )
+
+0811:             norm_res += (1. * n_t[i] / (np.sum(n_t) + n_c) * entropyH(pt_a_i))
+
    __pyx_t_1 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    if (unlikely(__pyx_v_n_t == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 811, __pyx_L1_error)
+    }
+    __pyx_t_2 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_12))) {
+      __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_12);
+      if (likely(__pyx_t_4)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_12);
+        __Pyx_INCREF(__pyx_t_4);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_12, function);
+      }
+    }
+    __pyx_t_3 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_12, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_12, __pyx_v_n_t);
+    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __pyx_t_12 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __pyx_t_4 = PyNumber_Add(__pyx_t_3, __pyx_t_12); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __pyx_t_12 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_v_pt_a_i); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
+    __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_6, NULL); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
+    __pyx_t_4 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_2 = PyNumber_Multiply(__pyx_t_12, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = PyNumber_InPlaceAdd(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_4); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_v_norm_res = __pyx_t_5;
+  }
+
 0812:         # Normalization Part 4
+
+0813:         norm_res += 1. * n_c/(np.sum(n_t) + n_c) * entropyH(pc_a)
+
  __pyx_t_4 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_2 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_np); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_12, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_t_12 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_12)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_12);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_12) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_12, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_12 = PyNumber_Add(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_12); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_v_pc_a, NULL); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 813, __pyx_L1_error)
+  __pyx_t_12 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __pyx_t_2 = PyNumber_Multiply(__pyx_t_3, __pyx_t_12); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_t_12 = PyNumber_InPlaceAdd(__pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_12); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 813, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_v_norm_res = __pyx_t_5;
+
 0814:
+
 0815:         # Normalization Part 5
+
+0816:         norm_res += 0.5
+
  __pyx_v_norm_res = (__pyx_v_norm_res + 0.5);
+
+0817:         return norm_res
+
  __Pyx_XDECREF(__pyx_r);
+  __pyx_t_12 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 817, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __pyx_r = __pyx_t_12;
+  __pyx_t_12 = 0;
+  goto __pyx_L0;
+
 0818:
+
+0819:     def tree_node_summary(self, treatment_idx, y, min_samples_treatment=10, n_reg=100, parentNodeSummary=None):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary[] = "\n        Tree node summary statistics.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        min_samples_treatment: int, optional (default=10)\n            The minimum number of samples required of the experiment group t be split at a leaf node.\n        n_reg :  int, optional (default=10)\n            The regularization parameter defined in Rzepakowski et al. 2012,\n            the weight (in terms of sample size) of the parent node influence\n            on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : list of list\n            The positive probabilities and sample sizes of each of the control and treatment groups\n            in the parent node.\n\n        Returns\n        -------\n        nodeSummary : list of list\n            The positive probabilities and sample sizes of each of the control and treatment groups\n            in the current node.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary = {"tree_node_summary", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_min_samples_treatment = 0;
+  PyObject *__pyx_v_n_reg = 0;
+  PyObject *__pyx_v_parentNodeSummary = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("tree_node_summary (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
+    PyObject* values[6] = {0,0,0,0,0,0};
+    values[3] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
+    values[5] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, 1); __PYX_ERR(0, 819, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, 2); __PYX_ERR(0, 819, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
+          if (value) { values[3] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
+          if (value) { values[5] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "tree_node_summary") < 0)) __PYX_ERR(0, 819, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_treatment_idx = values[1];
+    __pyx_v_y = values[2];
+    __pyx_v_min_samples_treatment = values[3];
+    __pyx_v_n_reg = values[4];
+    __pyx_v_parentNodeSummary = values[5];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 819, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.tree_node_summary", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
+  PyObject *__pyx_v_counts = NULL;
+  PyObject *__pyx_v_nodeSummary = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_count = NULL;
+  PyObject *__pyx_v_n_pos = NULL;
+  PyObject *__pyx_v_n = NULL;
+  PyObject *__pyx_v_p = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("tree_node_summary", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.tree_node_summary", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_counts);
+  __Pyx_XDECREF(__pyx_v_nodeSummary);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_count);
+  __Pyx_XDECREF(__pyx_v_n_pos);
+  __Pyx_XDECREF(__pyx_v_n);
+  __Pyx_XDECREF(__pyx_v_p);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__53 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_counts, __pyx_n_s_nodeSummary, __pyx_n_s_i, __pyx_n_s_count, __pyx_n_s_n_pos, __pyx_n_s_n, __pyx_n_s_p); if (unlikely(!__pyx_tuple__53)) __PYX_ERR(0, 819, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__53);
+  __Pyx_GIVEREF(__pyx_tuple__53);
+  __pyx_codeobj__54 = (PyObject*)__Pyx_PyCode_New(6, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__53, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_tree_node_summary, 819, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__54)) __PYX_ERR(0, 819, __pyx_L1_error)
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary, 0, __pyx_n_s_UpliftTreeClassifier_tree_node_s, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__54)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 819, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__55);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_tree_node_summary, __pyx_t_5) < 0) __PYX_ERR(0, 819, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_tuple__55 = PyTuple_Pack(3, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__55)) __PYX_ERR(0, 819, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__55);
+  __Pyx_GIVEREF(__pyx_tuple__55);
+
 0820:         '''
+
 0821:         Tree node summary statistics.
+
 0822:
+
 0823:         Args
+
 0824:         ----
+
 0825:         treatment_idx : array-like, shape = [num_samples]
+
 0826:             An array containing the treatment group index for each unit.
+
 0827:         y : array-like, shape = [num_samples]
+
 0828:             An array containing the outcome of interest for each unit.
+
 0829:         min_samples_treatment: int, optional (default=10)
+
 0830:             The minimum number of samples required of the experiment group t be split at a leaf node.
+
 0831:         n_reg :  int, optional (default=10)
+
 0832:             The regularization parameter defined in Rzepakowski et al. 2012,
+
 0833:             the weight (in terms of sample size) of the parent node influence
+
 0834:             on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
+
 0835:         parentNodeSummary : list of list
+
 0836:             The positive probabilities and sample sizes of each of the control and treatment groups
+
 0837:             in the parent node.
+
 0838:
+
 0839:         Returns
+
 0840:         -------
+
 0841:         nodeSummary : list of list
+
 0842:             The positive probabilities and sample sizes of each of the control and treatment groups
+
 0843:             in the current node.
+
 0844:         '''
+
 0845:         # counts: [[N(Y=0, T=0), N(Y=1, T=0)], [N(Y=0, T=1), N(Y=1, T=1)], ...]
+
+0846:         counts = self.group_uniqueCounts(treatment_idx, y)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_group_uniqueCounts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 846, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 846, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_treatment_idx);
+    __Pyx_GIVEREF(__pyx_v_treatment_idx);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_treatment_idx);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_counts = __pyx_t_1;
+  __pyx_t_1 = 0;
+
 0847:
+
 0848:         # nodeSummary: [[P(Y=1|T=0), N(T=0)], [P(Y=1|T=1), N(T=1)], ...]
+
+0849:         nodeSummary = []
+
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 849, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_nodeSummary = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
 0850:         # Iterate the control and treatment groups
+
+0851:         for i, count in enumerate(counts):
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_t_1 = __pyx_int_0;
+  if (likely(PyList_CheckExact(__pyx_v_counts)) || PyTuple_CheckExact(__pyx_v_counts)) {
+    __pyx_t_2 = __pyx_v_counts; __Pyx_INCREF(__pyx_t_2); __pyx_t_6 = 0;
+    __pyx_t_7 = NULL;
+  } else {
+    __pyx_t_6 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_counts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 851, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_7 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 851, __pyx_L1_error)
+  }
+  for (;;) {
+    if (likely(!__pyx_t_7)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 851, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      } else {
+        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 851, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      }
+    } else {
+      __pyx_t_5 = __pyx_t_7(__pyx_t_2);
+      if (unlikely(!__pyx_t_5)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 851, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_5);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_count, __pyx_t_5);
+    __pyx_t_5 = 0;
+    __Pyx_INCREF(__pyx_t_1);
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
+    __pyx_t_5 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_1);
+    __pyx_t_1 = __pyx_t_5;
+    __pyx_t_5 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0852:             n_pos = count[1]
+
    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_count, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 852, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_5);
+    __pyx_t_5 = 0;
+
+0853:             n = count[0] + n_pos
+
    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_count, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 853, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_3 = PyNumber_Add(__pyx_t_5, __pyx_v_n_pos); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 853, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_n, __pyx_t_3);
+    __pyx_t_3 = 0;
+
+0854:             if parentNodeSummary is None:
+
    __pyx_t_8 = (__pyx_v_parentNodeSummary == Py_None);
+    __pyx_t_9 = (__pyx_t_8 != 0);
+    if (__pyx_t_9) {
+/* … */
+      goto __pyx_L5;
+    }
+
+0855:                 p = n_pos / n if n > 0 else 0.
+
      __pyx_t_5 = PyObject_RichCompare(__pyx_v_n, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 855, __pyx_L1_error)
+      __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 855, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (__pyx_t_9) {
+        __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_v_n_pos, __pyx_v_n); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 855, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __pyx_t_3 = __pyx_t_5;
+        __pyx_t_5 = 0;
+      } else {
+        __Pyx_INCREF(__pyx_float_0_);
+        __pyx_t_3 = __pyx_float_0_;
+      }
+      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_3);
+      __pyx_t_3 = 0;
+
+0856:             elif n > min_samples_treatment:
+
    __pyx_t_3 = PyObject_RichCompare(__pyx_v_n, __pyx_v_min_samples_treatment, Py_GT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 856, __pyx_L1_error)
+    __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 856, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if (__pyx_t_9) {
+/* … */
+      goto __pyx_L5;
+    }
+
+0857:                 p = (n_pos + parentNodeSummary[i][0] * n_reg) / (n + n_reg)
+
      __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_parentNodeSummary, __pyx_v_i); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_3, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_3 = PyNumber_Multiply(__pyx_t_5, __pyx_v_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      __pyx_t_5 = PyNumber_Add(__pyx_v_n_pos, __pyx_t_3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_3 = PyNumber_Add(__pyx_v_n, __pyx_v_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 857, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_10);
+      __pyx_t_10 = 0;
+
 0858:             else:
+
+0859:                 p = parentNodeSummary[i][0]
+
    /*else*/ {
+      __pyx_t_10 = __Pyx_PyObject_GetItem(__pyx_v_parentNodeSummary, __pyx_v_i); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 859, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 859, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_3);
+      __pyx_t_3 = 0;
+    }
+    __pyx_L5:;
+
 0860:
+
+0861:             nodeSummary.append([p, n])
+
    __pyx_t_3 = PyList_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 861, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_INCREF(__pyx_v_p);
+    __Pyx_GIVEREF(__pyx_v_p);
+    PyList_SET_ITEM(__pyx_t_3, 0, __pyx_v_p);
+    __Pyx_INCREF(__pyx_v_n);
+    __Pyx_GIVEREF(__pyx_v_n);
+    PyList_SET_ITEM(__pyx_t_3, 1, __pyx_v_n);
+    __pyx_t_11 = __Pyx_PyList_Append(__pyx_v_nodeSummary, __pyx_t_3); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 861, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
 0862:
+
+0863:         return nodeSummary
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_nodeSummary);
+  __pyx_r = __pyx_v_nodeSummary;
+  goto __pyx_L0;
+
 0864:
+
+0865:     def uplift_classification_results(self, treatment_idx, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results[] = "\n        Classification probability for each treatment in the tree node.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        res : list of list\n            The positive probabilities P(Y = 1) of each of the control and treatment groups\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results = {"uplift_classification_results", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("uplift_classification_results (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,0};
+    PyObject* values[3] = {0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, 1); __PYX_ERR(0, 865, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, 2); __PYX_ERR(0, 865, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "uplift_classification_results") < 0)) __PYX_ERR(0, 865, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_treatment_idx = values[1];
+    __pyx_v_y = values[2];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 865, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.uplift_classification_results", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y) {
+  PyObject *__pyx_v_counts = NULL;
+  PyObject *__pyx_v_res = NULL;
+  PyObject *__pyx_v_count = NULL;
+  PyObject *__pyx_v_n_pos = NULL;
+  PyObject *__pyx_v_n = NULL;
+  PyObject *__pyx_v_p = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("uplift_classification_results", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.uplift_classification_results", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_counts);
+  __Pyx_XDECREF(__pyx_v_res);
+  __Pyx_XDECREF(__pyx_v_count);
+  __Pyx_XDECREF(__pyx_v_n_pos);
+  __Pyx_XDECREF(__pyx_v_n);
+  __Pyx_XDECREF(__pyx_v_p);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__56 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_counts, __pyx_n_s_res, __pyx_n_s_count, __pyx_n_s_n_pos, __pyx_n_s_n, __pyx_n_s_p); if (unlikely(!__pyx_tuple__56)) __PYX_ERR(0, 865, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__56);
+  __Pyx_GIVEREF(__pyx_tuple__56);
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results, 0, __pyx_n_s_UpliftTreeClassifier_uplift_clas, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__57)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 865, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_uplift_classification_results, __pyx_t_5) < 0) __PYX_ERR(0, 865, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_codeobj__57 = (PyObject*)__Pyx_PyCode_New(3, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__56, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_uplift_classification_results, 865, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__57)) __PYX_ERR(0, 865, __pyx_L1_error)
+
 0866:         '''
+
 0867:         Classification probability for each treatment in the tree node.
+
 0868:
+
 0869:         Args
+
 0870:         ----
+
 0871:         treatment_idx : array-like, shape = [num_samples]
+
 0872:             An array containing the treatment group index for each unit.
+
 0873:         y : array-like, shape = [num_samples]
+
 0874:             An array containing the outcome of interest for each unit.
+
 0875:
+
 0876:         Returns
+
 0877:         -------
+
 0878:         res : list of list
+
 0879:             The positive probabilities P(Y = 1) of each of the control and treatment groups
+
 0880:         '''
+
 0881:         # counts: [[N(Y=0, T=0), N(Y=1, T=0)], [N(Y=0, T=1), N(Y=1, T=1)], ...]
+
+0882:         counts = self.group_uniqueCounts(treatment_idx, y)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_group_uniqueCounts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 882, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = NULL;
+  __pyx_t_4 = 0;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_4 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+  } else
+  #endif
+  {
+    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 882, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (__pyx_t_3) {
+      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
+    }
+    __Pyx_INCREF(__pyx_v_treatment_idx);
+    __Pyx_GIVEREF(__pyx_v_treatment_idx);
+    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_treatment_idx);
+    __Pyx_INCREF(__pyx_v_y);
+    __Pyx_GIVEREF(__pyx_v_y);
+    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_counts = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+0883:         res = []
+
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 883, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_res = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
+0884:         for count in counts:
+
  if (likely(PyList_CheckExact(__pyx_v_counts)) || PyTuple_CheckExact(__pyx_v_counts)) {
+    __pyx_t_1 = __pyx_v_counts; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
+    __pyx_t_7 = NULL;
+  } else {
+    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_counts); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 884, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 884, __pyx_L1_error)
+  }
+  for (;;) {
+    if (likely(!__pyx_t_7)) {
+      if (likely(PyList_CheckExact(__pyx_t_1))) {
+        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 884, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 884, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      } else {
+        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 884, __pyx_L1_error)
+        #else
+        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 884, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_2);
+        #endif
+      }
+    } else {
+      __pyx_t_2 = __pyx_t_7(__pyx_t_1);
+      if (unlikely(!__pyx_t_2)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 884, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_2);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_count, __pyx_t_2);
+    __pyx_t_2 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+0885:             n_pos = count[1]
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_count, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 885, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_2);
+    __pyx_t_2 = 0;
+
+0886:             n = count[0] + n_pos
+
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_count, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 886, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_5 = PyNumber_Add(__pyx_t_2, __pyx_v_n_pos); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 886, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_n, __pyx_t_5);
+    __pyx_t_5 = 0;
+
+0887:             p = n_pos / n if n > 0 else 0.
+
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_n, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
+    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 887, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    if (__pyx_t_8) {
+      __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_v_n_pos, __pyx_v_n); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_5 = __pyx_t_2;
+      __pyx_t_2 = 0;
+    } else {
+      __Pyx_INCREF(__pyx_float_0_);
+      __pyx_t_5 = __pyx_float_0_;
+    }
+    __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_5);
+    __pyx_t_5 = 0;
+
+0888:             res.append(p)
+
    __pyx_t_9 = __Pyx_PyList_Append(__pyx_v_res, __pyx_v_p); if (unlikely(__pyx_t_9 == ((int)-1))) __PYX_ERR(0, 888, __pyx_L1_error)
+
+0889:         return res
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_res);
+  __pyx_r = __pyx_v_res;
+  goto __pyx_L0;
+
 0890:
+
+0891:     def growDecisionTreeFrom(self, X, treatment_idx, y, max_depth=10,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom[] = "\n        Train the uplift decision tree.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group idx for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        max_depth: int, optional (default=10)\n            The maximum depth of the tree.\n        min_samples_leaf: int, optional (default=100)\n            The minimum number of samples required to be split at a leaf node.\n        depth : int, optional (default = 1)\n            The current depth.\n        min_samples_treatment: int, optional (default=10)\n            The minimum number of samples required of the experiment group to be split at a leaf node.\n        n_reg: int, optional (default=10)\n            The regularization parameter defined in Rzepakowski et al. 2012,\n            the weight (in terms of sample size) of the parent node influence\n            on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : dictionary, optional (default = None)\n            Node summary statistics of the parent tree node.\n\n        Returns\n        -------\n        object of DecisionTree class\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom = {"growDecisionTreeFrom", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment_idx = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_max_depth = 0;
+  PyObject *__pyx_v_min_samples_leaf = 0;
+  PyObject *__pyx_v_depth = 0;
+  PyObject *__pyx_v_min_samples_treatment = 0;
+  PyObject *__pyx_v_n_reg = 0;
+  PyObject *__pyx_v_parentNodeSummary = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("growDecisionTreeFrom (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_depth,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
+    PyObject* values[10] = {0,0,0,0,0,0,0,0,0,0};
+    values[4] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[5] = ((PyObject *)((PyObject *)__pyx_int_100));
+    values[6] = ((PyObject *)((PyObject *)__pyx_int_1));
+    values[7] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[8] = ((PyObject *)((PyObject *)__pyx_int_100));
+/* … */
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_depth, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
+  PyObject *__pyx_v_currentNodeSummary = NULL;
+  PyObject *__pyx_v_currentScore = NULL;
+  PyObject *__pyx_v_maxAbsDiff = NULL;
+  PyObject *__pyx_v_maxDiff = NULL;
+  PyObject *__pyx_v_bestTreatment = NULL;
+  PyObject *__pyx_v_suboptTreatment = NULL;
+  PyObject *__pyx_v_maxDiffTreatment = NULL;
+  PyObject *__pyx_v_maxDiffSign = NULL;
+  PyObject *__pyx_v_p_c = NULL;
+  PyObject *__pyx_v_n_c = NULL;
+  PyObject *__pyx_v_i_tr = NULL;
+  PyObject *__pyx_v_p_t = NULL;
+  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
+  PyObject *__pyx_v_diff = NULL;
+  PyObject *__pyx_v_n_t = NULL;
+  PyObject *__pyx_v_p_value = NULL;
+  PyObject *__pyx_v_upliftScore = NULL;
+  PyObject *__pyx_v_bestGain = NULL;
+  PyObject *__pyx_v_bestGainImp = NULL;
+  PyObject *__pyx_v_bestAttribute = NULL;
+  PyObject *__pyx_v_columnCount = NULL;
+  PyObject *__pyx_v_max_features = NULL;
+  PyObject *__pyx_v_col = NULL;
+  PyObject *__pyx_v_columnValues = NULL;
+  PyObject *__pyx_v_lsUnique = NULL;
+  PyObject *__pyx_v_lspercentile = NULL;
+  PyObject *__pyx_v_value = NULL;
+  PyObject *__pyx_v_X_l = NULL;
+  PyObject *__pyx_v_X_r = NULL;
+  PyObject *__pyx_v_w_l = NULL;
+  PyObject *__pyx_v_w_r = NULL;
+  PyObject *__pyx_v_y_l = NULL;
+  PyObject *__pyx_v_y_r = NULL;
+  double __pyx_v_p;
+  PyObject *__pyx_v_leftNodeSummary = NULL;
+  PyObject *__pyx_v_rightNodeSummary = NULL;
+  PyObject *__pyx_v_node_mst = NULL;
+  PyObject *__pyx_v_leftScore1 = NULL;
+  PyObject *__pyx_v_rightScore2 = NULL;
+  PyObject *__pyx_v_gain = NULL;
+  PyObject *__pyx_v_gain_for_imp = NULL;
+  PyObject *__pyx_v_n_c_left = NULL;
+  PyObject *__pyx_v_n_t_left = NULL;
+  PyObject *__pyx_v_norm_factor = NULL;
+  PyObject *__pyx_v_best_set_left = NULL;
+  PyObject *__pyx_v_best_set_right = NULL;
+  PyObject *__pyx_v_dcY = NULL;
+  PyObject *__pyx_v_i = NULL;
+  PyObject *__pyx_v_summary = NULL;
+  PyObject *__pyx_v_trueBranch = NULL;
+  PyObject *__pyx_v_falseBranch = NULL;
+  PyObject *__pyx_v_stat = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("growDecisionTreeFrom", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_XDECREF(__pyx_t_11);
+  __Pyx_XDECREF(__pyx_t_12);
+  __Pyx_XDECREF(__pyx_t_16);
+  __Pyx_XDECREF(__pyx_t_17);
+  __Pyx_XDECREF(__pyx_t_21);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.growDecisionTreeFrom", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
+  __Pyx_XDECREF(__pyx_v_currentScore);
+  __Pyx_XDECREF(__pyx_v_maxAbsDiff);
+  __Pyx_XDECREF(__pyx_v_maxDiff);
+  __Pyx_XDECREF(__pyx_v_bestTreatment);
+  __Pyx_XDECREF(__pyx_v_suboptTreatment);
+  __Pyx_XDECREF(__pyx_v_maxDiffTreatment);
+  __Pyx_XDECREF(__pyx_v_maxDiffSign);
+  __Pyx_XDECREF(__pyx_v_p_c);
+  __Pyx_XDECREF(__pyx_v_n_c);
+  __Pyx_XDECREF(__pyx_v_i_tr);
+  __Pyx_XDECREF(__pyx_v_p_t);
+  __Pyx_XDECREF(__pyx_v__);
+  __Pyx_XDECREF(__pyx_v_diff);
+  __Pyx_XDECREF(__pyx_v_n_t);
+  __Pyx_XDECREF(__pyx_v_p_value);
+  __Pyx_XDECREF(__pyx_v_upliftScore);
+  __Pyx_XDECREF(__pyx_v_bestGain);
+  __Pyx_XDECREF(__pyx_v_bestGainImp);
+  __Pyx_XDECREF(__pyx_v_bestAttribute);
+  __Pyx_XDECREF(__pyx_v_columnCount);
+  __Pyx_XDECREF(__pyx_v_max_features);
+  __Pyx_XDECREF(__pyx_v_col);
+  __Pyx_XDECREF(__pyx_v_columnValues);
+  __Pyx_XDECREF(__pyx_v_lsUnique);
+  __Pyx_XDECREF(__pyx_v_lspercentile);
+  __Pyx_XDECREF(__pyx_v_value);
+  __Pyx_XDECREF(__pyx_v_X_l);
+  __Pyx_XDECREF(__pyx_v_X_r);
+  __Pyx_XDECREF(__pyx_v_w_l);
+  __Pyx_XDECREF(__pyx_v_w_r);
+  __Pyx_XDECREF(__pyx_v_y_l);
+  __Pyx_XDECREF(__pyx_v_y_r);
+  __Pyx_XDECREF(__pyx_v_leftNodeSummary);
+  __Pyx_XDECREF(__pyx_v_rightNodeSummary);
+  __Pyx_XDECREF(__pyx_v_node_mst);
+  __Pyx_XDECREF(__pyx_v_leftScore1);
+  __Pyx_XDECREF(__pyx_v_rightScore2);
+  __Pyx_XDECREF(__pyx_v_gain);
+  __Pyx_XDECREF(__pyx_v_gain_for_imp);
+  __Pyx_XDECREF(__pyx_v_n_c_left);
+  __Pyx_XDECREF(__pyx_v_n_t_left);
+  __Pyx_XDECREF(__pyx_v_norm_factor);
+  __Pyx_XDECREF(__pyx_v_best_set_left);
+  __Pyx_XDECREF(__pyx_v_best_set_right);
+  __Pyx_XDECREF(__pyx_v_dcY);
+  __Pyx_XDECREF(__pyx_v_i);
+  __Pyx_XDECREF(__pyx_v_summary);
+  __Pyx_XDECREF(__pyx_v_trueBranch);
+  __Pyx_XDECREF(__pyx_v_falseBranch);
+  __Pyx_XDECREF(__pyx_v_stat);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__58 = PyTuple_Pack(63, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_depth, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_currentNodeSummary, __pyx_n_s_currentScore, __pyx_n_s_maxAbsDiff, __pyx_n_s_maxDiff, __pyx_n_s_bestTreatment, __pyx_n_s_suboptTreatment, __pyx_n_s_maxDiffTreatment, __pyx_n_s_maxDiffSign, __pyx_n_s_p_c, __pyx_n_s_n_c, __pyx_n_s_i_tr, __pyx_n_s_p_t, __pyx_n_s__34, __pyx_n_s_diff, __pyx_n_s_n_t, __pyx_n_s_p_value, __pyx_n_s_upliftScore, __pyx_n_s_bestGain, __pyx_n_s_bestGainImp, __pyx_n_s_bestAttribute, __pyx_n_s_columnCount, __pyx_n_s_max_features, __pyx_n_s_col, __pyx_n_s_columnValues, __pyx_n_s_lsUnique, __pyx_n_s_lspercentile, __pyx_n_s_value, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_p, __pyx_n_s_leftNodeSummary, __pyx_n_s_rightNodeSummary, __pyx_n_s_node_mst, __pyx_n_s_leftScore1, __pyx_n_s_rightScore2, __pyx_n_s_gain, __pyx_n_s_gain_for_imp, __pyx_n_s_n_c_left, __pyx_n_s_n_t_left, __pyx_n_s_norm_factor, __pyx_n_s_best_set_left, __pyx_n_s_best_set_right, __pyx_n_s_dcY, __pyx_n_s_i, __pyx_n_s_summary, __pyx_n_s_trueBranch, __pyx_n_s_falseBranch, __pyx_n_s_stat, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__58)) __PYX_ERR(0, 891, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__58);
+  __Pyx_GIVEREF(__pyx_tuple__58);
+  __pyx_codeobj__59 = (PyObject*)__Pyx_PyCode_New(10, 0, 63, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__58, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_growDecisionTreeFrom, 891, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__59)) __PYX_ERR(0, 891, __pyx_L1_error)
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom, 0, __pyx_n_s_UpliftTreeClassifier_growDecisio, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__59)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 891, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__60);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_growDecisionTreeFrom, __pyx_t_5) < 0) __PYX_ERR(0, 891, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_tuple__60 = PyTuple_Pack(6, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_1), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__60)) __PYX_ERR(0, 891, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__60);
+  __Pyx_GIVEREF(__pyx_tuple__60);
+
 0892:                              min_samples_leaf=100, depth=1,
+
 0893:                              min_samples_treatment=10, n_reg=100,
+
+0894:                              parentNodeSummary=None):
+
    values[9] = ((PyObject *)((PyObject *)Py_None));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 1); __PYX_ERR(0, 891, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 2); __PYX_ERR(0, 891, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 3); __PYX_ERR(0, 891, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_depth);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
+          if (value) { values[7] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  8:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
+          if (value) { values[8] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  9:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
+          if (value) { values[9] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "growDecisionTreeFrom") < 0)) __PYX_ERR(0, 891, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment_idx = values[2];
+    __pyx_v_y = values[3];
+    __pyx_v_max_depth = values[4];
+    __pyx_v_min_samples_leaf = values[5];
+    __pyx_v_depth = values[6];
+    __pyx_v_min_samples_treatment = values[7];
+    __pyx_v_n_reg = values[8];
+    __pyx_v_parentNodeSummary = values[9];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 891, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.growDecisionTreeFrom", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_depth, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
+
 0895:         '''
+
 0896:         Train the uplift decision tree.
+
 0897:
+
 0898:         Args
+
 0899:         ----
+
 0900:         X : ndarray, shape = [num_samples, num_features]
+
 0901:             An ndarray of the covariates used to train the uplift model.
+
 0902:         treatment_idx : array-like, shape = [num_samples]
+
 0903:             An array containing the treatment group idx for each unit.
+
 0904:         y : array-like, shape = [num_samples]
+
 0905:             An array containing the outcome of interest for each unit.
+
 0906:         max_depth: int, optional (default=10)
+
 0907:             The maximum depth of the tree.
+
 0908:         min_samples_leaf: int, optional (default=100)
+
 0909:             The minimum number of samples required to be split at a leaf node.
+
 0910:         depth : int, optional (default = 1)
+
 0911:             The current depth.
+
 0912:         min_samples_treatment: int, optional (default=10)
+
 0913:             The minimum number of samples required of the experiment group to be split at a leaf node.
+
 0914:         n_reg: int, optional (default=10)
+
 0915:             The regularization parameter defined in Rzepakowski et al. 2012,
+
 0916:             the weight (in terms of sample size) of the parent node influence
+
 0917:             on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
+
 0918:         parentNodeSummary : dictionary, optional (default = None)
+
 0919:             Node summary statistics of the parent tree node.
+
 0920:
+
 0921:         Returns
+
 0922:         -------
+
 0923:         object of DecisionTree class
+
 0924:         '''
+
 0925:
+
+0926:         if len(X) == 0:
+
  __pyx_t_1 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_1 == ((Py_ssize_t)-1))) __PYX_ERR(0, 926, __pyx_L1_error)
+  __pyx_t_2 = ((__pyx_t_1 == 0) != 0);
+  if (__pyx_t_2) {
+/* … */
+  }
+
+0927:             return DecisionTree(classes_=self.classes_)
+
    __Pyx_XDECREF(__pyx_r);
+    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 927, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 927, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 927, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_classes, __pyx_t_5) < 0) __PYX_ERR(0, 927, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 927, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_r = __pyx_t_5;
+    __pyx_t_5 = 0;
+    goto __pyx_L0;
+
 0928:
+
 0929:         # Current node summary: [P(Y=1|T), N(T)]
+
+0930:         currentNodeSummary = self.tree_node_summary(treatment_idx, y,
+
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 930, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 930, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_INCREF(__pyx_v_treatment_idx);
+  __Pyx_GIVEREF(__pyx_v_treatment_idx);
+  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_treatment_idx);
+  __Pyx_INCREF(__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_v_y);
+  PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y);
+/* … */
+  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_4, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 930, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_v_currentNodeSummary = __pyx_t_6;
+  __pyx_t_6 = 0;
+
+0931:                                                     min_samples_treatment=min_samples_treatment,
+
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 931, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
+
+0932:                                                     n_reg=n_reg,
+
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
+
+0933:                                                     parentNodeSummary=parentNodeSummary)
+
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, __pyx_v_parentNodeSummary) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
+
+0934:         currentScore = self.evaluationFunction(currentNodeSummary)
+
  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 934, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_6 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_currentNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_currentNodeSummary);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 934, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_v_currentScore = __pyx_t_6;
+  __pyx_t_6 = 0;
+
 0935:
+
 0936:         # Prune Stats
+
+0937:         maxAbsDiff = 0
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_maxAbsDiff = __pyx_int_0;
+
+0938:         maxDiff = -1.
+
  __Pyx_INCREF(__pyx_float_neg_1_);
+  __pyx_v_maxDiff = __pyx_float_neg_1_;
+
+0939:         bestTreatment = 0       # treatment index for the control group
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_bestTreatment = __pyx_int_0;
+
+0940:         suboptTreatment = 0     # treatment index for the control group
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_suboptTreatment = __pyx_int_0;
+
+0941:         maxDiffTreatment = 0    # treatment index for the control group
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_maxDiffTreatment = __pyx_int_0;
+
+0942:         maxDiffSign = 0
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_v_maxDiffSign = __pyx_int_0;
+
+0943:         p_c, n_c = currentNodeSummary[0]
+
  __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 943, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
+    PyObject* sequence = __pyx_t_6;
+    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+    if (unlikely(size != 2)) {
+      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+      __PYX_ERR(0, 943, __pyx_L1_error)
+    }
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    if (likely(PyTuple_CheckExact(sequence))) {
+      __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
+      __pyx_t_4 = PyTuple_GET_ITEM(sequence, 1); 
+    } else {
+      __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
+      __pyx_t_4 = PyList_GET_ITEM(sequence, 1); 
+    }
+    __Pyx_INCREF(__pyx_t_3);
+    __Pyx_INCREF(__pyx_t_4);
+    #else
+    __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_4 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    #endif
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  } else {
+    Py_ssize_t index = -1;
+    __pyx_t_5 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 943, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_7 = Py_TYPE(__pyx_t_5)->tp_iternext;
+    index = 0; __pyx_t_3 = __pyx_t_7(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L4_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_3);
+    index = 1; __pyx_t_4 = __pyx_t_7(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L4_unpacking_failed;
+    __Pyx_GOTREF(__pyx_t_4);
+    if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_5), 2) < 0) __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_t_7 = NULL;
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    goto __pyx_L5_unpacking_done;
+    __pyx_L4_unpacking_failed:;
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_7 = NULL;
+    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+    __PYX_ERR(0, 943, __pyx_L1_error)
+    __pyx_L5_unpacking_done:;
+  }
+  __pyx_v_p_c = __pyx_t_3;
+  __pyx_t_3 = 0;
+  __pyx_v_n_c = __pyx_t_4;
+  __pyx_t_4 = 0;
+
+0944:         for i_tr in range(1, self.n_class):
+
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 944, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_INCREF(__pyx_int_1);
+  __Pyx_GIVEREF(__pyx_int_1);
+  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_int_1);
+  __Pyx_GIVEREF(__pyx_t_6);
+  PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_t_6);
+  __pyx_t_6 = 0;
+  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_4, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_6)) || PyTuple_CheckExact(__pyx_t_6)) {
+    __pyx_t_4 = __pyx_t_6; __Pyx_INCREF(__pyx_t_4); __pyx_t_1 = 0;
+    __pyx_t_8 = NULL;
+  } else {
+    __pyx_t_1 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 944, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_8 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 944, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_8)) {
+      if (likely(PyList_CheckExact(__pyx_t_4))) {
+        if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_4)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_6 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_1); __Pyx_INCREF(__pyx_t_6); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 944, __pyx_L1_error)
+        #else
+        __pyx_t_6 = PySequence_ITEM(__pyx_t_4, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_6);
+        #endif
+      } else {
+        if (__pyx_t_1 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_6 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_1); __Pyx_INCREF(__pyx_t_6); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 944, __pyx_L1_error)
+        #else
+        __pyx_t_6 = PySequence_ITEM(__pyx_t_4, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_6);
+        #endif
+      }
+    } else {
+      __pyx_t_6 = __pyx_t_8(__pyx_t_4);
+      if (unlikely(!__pyx_t_6)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 944, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_6);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_i_tr, __pyx_t_6);
+    __pyx_t_6 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0945:             p_t, _ = currentNodeSummary[i_tr]
+
    __pyx_t_6 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_i_tr); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 945, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
+      PyObject* sequence = __pyx_t_6;
+      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+      if (unlikely(size != 2)) {
+        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+        __PYX_ERR(0, 945, __pyx_L1_error)
+      }
+      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+      if (likely(PyTuple_CheckExact(sequence))) {
+        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
+        __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
+      } else {
+        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
+        __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
+      }
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_5);
+      #else
+      __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 945, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 945, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      #endif
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    } else {
+      Py_ssize_t index = -1;
+      __pyx_t_9 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 945, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __pyx_t_7 = Py_TYPE(__pyx_t_9)->tp_iternext;
+      index = 0; __pyx_t_3 = __pyx_t_7(__pyx_t_9); if (unlikely(!__pyx_t_3)) goto __pyx_L8_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_3);
+      index = 1; __pyx_t_5 = __pyx_t_7(__pyx_t_9); if (unlikely(!__pyx_t_5)) goto __pyx_L8_unpacking_failed;
+      __Pyx_GOTREF(__pyx_t_5);
+      if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_9), 2) < 0) __PYX_ERR(0, 945, __pyx_L1_error)
+      __pyx_t_7 = NULL;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      goto __pyx_L9_unpacking_done;
+      __pyx_L8_unpacking_failed:;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __pyx_t_7 = NULL;
+      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+      __PYX_ERR(0, 945, __pyx_L1_error)
+      __pyx_L9_unpacking_done:;
+    }
+    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_5);
+    __pyx_t_5 = 0;
+
 0946:             # P(Y|T=t) - P(Y|T=0)
+
+0947:             diff = p_t - p_c
+
    __pyx_t_6 = PyNumber_Subtract(__pyx_v_p_t, __pyx_v_p_c); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 947, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_XDECREF_SET(__pyx_v_diff, __pyx_t_6);
+    __pyx_t_6 = 0;
+
+0948:             if abs(diff) >= maxAbsDiff:
+
    __pyx_t_6 = __Pyx_PyNumber_Absolute(__pyx_v_diff); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 948, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_5 = PyObject_RichCompare(__pyx_t_6, __pyx_v_maxAbsDiff, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 948, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 948, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (__pyx_t_2) {
+/* … */
+    }
+
+0949:                 maxDiffTreatment = i_tr
+
      __Pyx_INCREF(__pyx_v_i_tr);
+      __Pyx_DECREF_SET(__pyx_v_maxDiffTreatment, __pyx_v_i_tr);
+
+0950:                 maxDiffSign = np.sign(diff)
+
      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 950, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_6);
+      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_sign); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 950, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __pyx_t_6 = NULL;
+      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+        __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_3);
+        if (likely(__pyx_t_6)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+          __Pyx_INCREF(__pyx_t_6);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_3, function);
+        }
+      }
+      __pyx_t_5 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_6, __pyx_v_diff) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_diff);
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 950, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_DECREF_SET(__pyx_v_maxDiffSign, __pyx_t_5);
+      __pyx_t_5 = 0;
+
+0951:                 maxAbsDiff = abs(diff)
+
      __pyx_t_5 = __Pyx_PyNumber_Absolute(__pyx_v_diff); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 951, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __Pyx_DECREF_SET(__pyx_v_maxAbsDiff, __pyx_t_5);
+      __pyx_t_5 = 0;
+
+0952:             if diff >= maxDiff:
+
    __pyx_t_5 = PyObject_RichCompare(__pyx_v_diff, __pyx_v_maxDiff, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 952, __pyx_L1_error)
+    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 952, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (__pyx_t_2) {
+/* … */
+    }
+
+0953:                 maxDiff = diff
+
      __Pyx_INCREF(__pyx_v_diff);
+      __Pyx_DECREF_SET(__pyx_v_maxDiff, __pyx_v_diff);
+
+0954:                 suboptTreatment = i_tr
+
      __Pyx_INCREF(__pyx_v_i_tr);
+      __Pyx_DECREF_SET(__pyx_v_suboptTreatment, __pyx_v_i_tr);
+
+0955:                 if diff > 0:
+
      __pyx_t_5 = PyObject_RichCompare(__pyx_v_diff, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 955, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 955, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (__pyx_t_2) {
+/* … */
+      }
+
+0956:                     bestTreatment = i_tr
+
        __Pyx_INCREF(__pyx_v_i_tr);
+        __Pyx_DECREF_SET(__pyx_v_bestTreatment, __pyx_v_i_tr);
+
+0957:         if maxDiff > 0:
+
  __pyx_t_4 = PyObject_RichCompare(__pyx_v_maxDiff, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 957, __pyx_L1_error)
+  __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 957, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (__pyx_t_2) {
+/* … */
+    goto __pyx_L13;
+  }
+
+0958:             p_t = currentNodeSummary[bestTreatment][0]
+
    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_bestTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 958, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 958, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_5);
+    __pyx_t_5 = 0;
+
+0959:             n_t = currentNodeSummary[bestTreatment][1]
+
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_bestTreatment); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 959, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_5, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 959, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_v_n_t = __pyx_t_4;
+    __pyx_t_4 = 0;
+
 0960:         else:
+
+0961:             p_t = currentNodeSummary[suboptTreatment][0]
+
  /*else*/ {
+    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_suboptTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 961, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 961, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_5);
+    __pyx_t_5 = 0;
+
+0962:             n_t = currentNodeSummary[suboptTreatment][1]
+
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_suboptTreatment); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 962, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_5, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 962, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_v_n_t = __pyx_t_4;
+    __pyx_t_4 = 0;
+  }
+  __pyx_L13:;
+
 0963:
+
+0964:         p_value = (1. - stats.norm.cdf(abs(p_c - p_t) / np.sqrt(p_t * (1 - p_t) / n_t + p_c * (1 - p_c) / n_c))) * 2
+
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_stats); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_norm); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_cdf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = PyNumber_Subtract(__pyx_v_p_c, __pyx_v_p_t); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_6 = __Pyx_PyNumber_Absolute(__pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_sqrt); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_p_t, 1, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __pyx_t_11 = PyNumber_Multiply(__pyx_v_p_t, __pyx_t_9); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_9 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_v_n_t); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __pyx_t_11 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_p_c, 1, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __pyx_t_12 = PyNumber_Multiply(__pyx_v_p_c, __pyx_t_11); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __pyx_t_11 = __Pyx_PyNumber_Divide(__pyx_t_12, __pyx_v_n_c); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_11);
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_t_12 = PyNumber_Add(__pyx_t_9, __pyx_t_11); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __pyx_t_11 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_10))) {
+    __pyx_t_11 = PyMethod_GET_SELF(__pyx_t_10);
+    if (likely(__pyx_t_11)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+      __Pyx_INCREF(__pyx_t_11);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_10, function);
+    }
+  }
+  __pyx_t_3 = (__pyx_t_11) ? __Pyx_PyObject_Call2Args(__pyx_t_10, __pyx_t_11, __pyx_t_12) : __Pyx_PyObject_CallOneArg(__pyx_t_10, __pyx_t_12);
+  __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_6, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
+    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_5);
+    if (likely(__pyx_t_3)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+      __Pyx_INCREF(__pyx_t_3);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_5, function);
+    }
+  }
+  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_3, __pyx_t_10) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_10);
+  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = __Pyx_PyFloat_SubtractCObj(__pyx_float_1_, __pyx_t_4, 1., 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_4 = PyNumber_Multiply(__pyx_t_5, __pyx_int_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 964, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_p_value = __pyx_t_4;
+  __pyx_t_4 = 0;
+
+0965:         upliftScore = [maxDiff, p_value]
+
  __pyx_t_4 = PyList_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 965, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_INCREF(__pyx_v_maxDiff);
+  __Pyx_GIVEREF(__pyx_v_maxDiff);
+  PyList_SET_ITEM(__pyx_t_4, 0, __pyx_v_maxDiff);
+  __Pyx_INCREF(__pyx_v_p_value);
+  __Pyx_GIVEREF(__pyx_v_p_value);
+  PyList_SET_ITEM(__pyx_t_4, 1, __pyx_v_p_value);
+  __pyx_v_upliftScore = ((PyObject*)__pyx_t_4);
+  __pyx_t_4 = 0;
+
 0966:
+
+0967:         bestGain = 0.0
+
  __Pyx_INCREF(__pyx_float_0_0);
+  __pyx_v_bestGain = __pyx_float_0_0;
+
+0968:         bestGainImp = 0.0
+
  __Pyx_INCREF(__pyx_float_0_0);
+  __pyx_v_bestGainImp = __pyx_float_0_0;
+
+0969:         bestAttribute = None
+
  __Pyx_INCREF(Py_None);
+  __pyx_v_bestAttribute = ((PyObject*)Py_None);
+
 0970:
+
 0971:         # last column is the result/target column, 2nd to the last is the treatment group
+
+0972:         columnCount = X.shape[1]
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 972, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 972, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_v_columnCount = __pyx_t_5;
+  __pyx_t_5 = 0;
+
+0973:         if (self.max_features and self.max_features > 0 and self.max_features <= columnCount):
+
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (__pyx_t_13) {
+  } else {
+    __pyx_t_2 = __pyx_t_13;
+    goto __pyx_L15_bool_binop_done;
+  }
+  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_4 = PyObject_RichCompare(__pyx_t_5, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (__pyx_t_13) {
+  } else {
+    __pyx_t_2 = __pyx_t_13;
+    goto __pyx_L15_bool_binop_done;
+  }
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_5 = PyObject_RichCompare(__pyx_t_4, __pyx_v_columnCount, Py_LE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_2 = __pyx_t_13;
+  __pyx_L15_bool_binop_done:;
+  if (__pyx_t_2) {
+/* … */
+    goto __pyx_L14;
+  }
+
+0974:             max_features = self.max_features
+
    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 974, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_v_max_features = __pyx_t_5;
+    __pyx_t_5 = 0;
+
 0975:         else:
+
+0976:             max_features = columnCount
+
  /*else*/ {
+    __Pyx_INCREF(__pyx_v_columnCount);
+    __pyx_v_max_features = __pyx_v_columnCount;
+  }
+  __pyx_L14:;
+
 0977:
+
+0978:         for col in list(self.random_state_.choice(a=range(columnCount), size=max_features, replace=False)):
+
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_choice); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_10 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_v_columnCount); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_a, __pyx_t_10) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_size, __pyx_v_max_features) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_replace, Py_False) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
+  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = PySequence_List(__pyx_t_10); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = __pyx_t_5; __Pyx_INCREF(__pyx_t_10); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  for (;;) {
+    if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_10)) break;
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    __pyx_t_5 = PyList_GET_ITEM(__pyx_t_10, __pyx_t_1); __Pyx_INCREF(__pyx_t_5); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 978, __pyx_L1_error)
+    #else
+    __pyx_t_5 = PySequence_ITEM(__pyx_t_10, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    #endif
+    __Pyx_XDECREF_SET(__pyx_v_col, __pyx_t_5);
+    __pyx_t_5 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
+0979:             columnValues = X[:, col]
+
    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 979, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_INCREF(__pyx_slice__7);
+    __Pyx_GIVEREF(__pyx_slice__7);
+    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_slice__7);
+    __Pyx_INCREF(__pyx_v_col);
+    __Pyx_GIVEREF(__pyx_v_col);
+    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_col);
+    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 979, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_columnValues, __pyx_t_4);
+    __pyx_t_4 = 0;
+
 0980:             # unique values
+
+0981:             lsUnique = np.unique(columnValues)
+
    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 981, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_unique); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 981, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_3);
+      if (likely(__pyx_t_5)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+        __Pyx_INCREF(__pyx_t_5);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_3, function);
+      }
+    }
+    __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_5, __pyx_v_columnValues) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_columnValues);
+    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 981, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_lsUnique, __pyx_t_4);
+    __pyx_t_4 = 0;
+
 0982:
+
+0983:             if np.issubdtype(lsUnique.dtype, np.number):
+
    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 983, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_issubdtype); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 983, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_lsUnique, __pyx_n_s_dtype); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 983, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __pyx_t_6 = NULL;
+    __pyx_t_14 = 0;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_5);
+      if (likely(__pyx_t_6)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+        __Pyx_INCREF(__pyx_t_6);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_5, function);
+        __pyx_t_14 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_5)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, ((PyObject *)__pyx_ptype_5numpy_number)};
+      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, ((PyObject *)__pyx_ptype_5numpy_number)};
+      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    } else
+    #endif
+    {
+      __pyx_t_12 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 983, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+      if (__pyx_t_6) {
+        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_6); __pyx_t_6 = NULL;
+      }
+      __Pyx_GIVEREF(__pyx_t_3);
+      PyTuple_SET_ITEM(__pyx_t_12, 0+__pyx_t_14, __pyx_t_3);
+      __Pyx_INCREF(((PyObject *)__pyx_ptype_5numpy_number));
+      __Pyx_GIVEREF(((PyObject *)__pyx_ptype_5numpy_number));
+      PyTuple_SET_ITEM(__pyx_t_12, 1+__pyx_t_14, ((PyObject *)__pyx_ptype_5numpy_number));
+      __pyx_t_3 = 0;
+      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_12, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 983, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (__pyx_t_2) {
+/* … */
+    }
+
+0984:                 if len(lsUnique) > 10:
+
      __pyx_t_15 = PyObject_Length(__pyx_v_lsUnique); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 984, __pyx_L1_error)
+      __pyx_t_2 = ((__pyx_t_15 > 10) != 0);
+      if (__pyx_t_2) {
+/* … */
+        goto __pyx_L21;
+      }
+
+0985:                     lspercentile = np.percentile(columnValues, [3, 5, 10, 20, 30, 50, 70, 80, 90, 95, 97])
+
        __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 985, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_percentile); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 985, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        __pyx_t_5 = PyList_New(11); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 985, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __Pyx_INCREF(__pyx_int_3);
+        __Pyx_GIVEREF(__pyx_int_3);
+        PyList_SET_ITEM(__pyx_t_5, 0, __pyx_int_3);
+        __Pyx_INCREF(__pyx_int_5);
+        __Pyx_GIVEREF(__pyx_int_5);
+        PyList_SET_ITEM(__pyx_t_5, 1, __pyx_int_5);
+        __Pyx_INCREF(__pyx_int_10);
+        __Pyx_GIVEREF(__pyx_int_10);
+        PyList_SET_ITEM(__pyx_t_5, 2, __pyx_int_10);
+        __Pyx_INCREF(__pyx_int_20);
+        __Pyx_GIVEREF(__pyx_int_20);
+        PyList_SET_ITEM(__pyx_t_5, 3, __pyx_int_20);
+        __Pyx_INCREF(__pyx_int_30);
+        __Pyx_GIVEREF(__pyx_int_30);
+        PyList_SET_ITEM(__pyx_t_5, 4, __pyx_int_30);
+        __Pyx_INCREF(__pyx_int_50);
+        __Pyx_GIVEREF(__pyx_int_50);
+        PyList_SET_ITEM(__pyx_t_5, 5, __pyx_int_50);
+        __Pyx_INCREF(__pyx_int_70);
+        __Pyx_GIVEREF(__pyx_int_70);
+        PyList_SET_ITEM(__pyx_t_5, 6, __pyx_int_70);
+        __Pyx_INCREF(__pyx_int_80);
+        __Pyx_GIVEREF(__pyx_int_80);
+        PyList_SET_ITEM(__pyx_t_5, 7, __pyx_int_80);
+        __Pyx_INCREF(__pyx_int_90);
+        __Pyx_GIVEREF(__pyx_int_90);
+        PyList_SET_ITEM(__pyx_t_5, 8, __pyx_int_90);
+        __Pyx_INCREF(__pyx_int_95);
+        __Pyx_GIVEREF(__pyx_int_95);
+        PyList_SET_ITEM(__pyx_t_5, 9, __pyx_int_95);
+        __Pyx_INCREF(__pyx_int_97);
+        __Pyx_GIVEREF(__pyx_int_97);
+        PyList_SET_ITEM(__pyx_t_5, 10, __pyx_int_97);
+        __pyx_t_3 = NULL;
+        __pyx_t_14 = 0;
+        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_12))) {
+          __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_12);
+          if (likely(__pyx_t_3)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_12);
+            __Pyx_INCREF(__pyx_t_3);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_12, function);
+            __pyx_t_14 = 1;
+          }
+        }
+        #if CYTHON_FAST_PYCALL
+        if (PyFunction_Check(__pyx_t_12)) {
+          PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_columnValues, __pyx_t_5};
+          __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_12, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        } else
+        #endif
+        #if CYTHON_FAST_PYCCALL
+        if (__Pyx_PyFastCFunction_Check(__pyx_t_12)) {
+          PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_columnValues, __pyx_t_5};
+          __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_12, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        } else
+        #endif
+        {
+          __pyx_t_6 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 985, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_6);
+          if (__pyx_t_3) {
+            __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
+          }
+          __Pyx_INCREF(__pyx_v_columnValues);
+          __Pyx_GIVEREF(__pyx_v_columnValues);
+          PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_14, __pyx_v_columnValues);
+          __Pyx_GIVEREF(__pyx_t_5);
+          PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_14, __pyx_t_5);
+          __pyx_t_5 = 0;
+          __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_t_6, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+        }
+        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_lspercentile, __pyx_t_4);
+        __pyx_t_4 = 0;
+
 0986:                 else:
+
+0987:                     lspercentile = np.percentile(lsUnique, [10, 50, 90])
+
      /*else*/ {
+        __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_np); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 987, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_12, __pyx_n_s_percentile); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 987, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_6);
+        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 987, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_INCREF(__pyx_int_10);
+        __Pyx_GIVEREF(__pyx_int_10);
+        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_int_10);
+        __Pyx_INCREF(__pyx_int_50);
+        __Pyx_GIVEREF(__pyx_int_50);
+        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_int_50);
+        __Pyx_INCREF(__pyx_int_90);
+        __Pyx_GIVEREF(__pyx_int_90);
+        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_int_90);
+        __pyx_t_5 = NULL;
+        __pyx_t_14 = 0;
+        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
+          __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
+          if (likely(__pyx_t_5)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+            __Pyx_INCREF(__pyx_t_5);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_6, function);
+            __pyx_t_14 = 1;
+          }
+        }
+        #if CYTHON_FAST_PYCALL
+        if (PyFunction_Check(__pyx_t_6)) {
+          PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_lsUnique, __pyx_t_12};
+          __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+        } else
+        #endif
+        #if CYTHON_FAST_PYCCALL
+        if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
+          PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_lsUnique, __pyx_t_12};
+          __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
+          __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+        } else
+        #endif
+        {
+          __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 987, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          if (__pyx_t_5) {
+            __Pyx_GIVEREF(__pyx_t_5); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_5); __pyx_t_5 = NULL;
+          }
+          __Pyx_INCREF(__pyx_v_lsUnique);
+          __Pyx_GIVEREF(__pyx_v_lsUnique);
+          PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_lsUnique);
+          __Pyx_GIVEREF(__pyx_t_12);
+          PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_t_12);
+          __pyx_t_12 = 0;
+          __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_3, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        }
+        __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_lspercentile, __pyx_t_4);
+        __pyx_t_4 = 0;
+      }
+      __pyx_L21:;
+
+0988:                 lsUnique = np.unique(lspercentile)
+
      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 988, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_6);
+      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_unique); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 988, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __pyx_t_6 = NULL;
+      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+        __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_3);
+        if (likely(__pyx_t_6)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+          __Pyx_INCREF(__pyx_t_6);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_3, function);
+        }
+      }
+      __pyx_t_4 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_6, __pyx_v_lspercentile) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_lspercentile);
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 988, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_DECREF_SET(__pyx_v_lsUnique, __pyx_t_4);
+      __pyx_t_4 = 0;
+
 0989:
+
+0990:             for value in lsUnique:
+
    if (likely(PyList_CheckExact(__pyx_v_lsUnique)) || PyTuple_CheckExact(__pyx_v_lsUnique)) {
+      __pyx_t_4 = __pyx_v_lsUnique; __Pyx_INCREF(__pyx_t_4); __pyx_t_15 = 0;
+      __pyx_t_8 = NULL;
+    } else {
+      __pyx_t_15 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_v_lsUnique); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 990, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_8 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 990, __pyx_L1_error)
+    }
+    for (;;) {
+      if (likely(!__pyx_t_8)) {
+        if (likely(PyList_CheckExact(__pyx_t_4))) {
+          if (__pyx_t_15 >= PyList_GET_SIZE(__pyx_t_4)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_15); __Pyx_INCREF(__pyx_t_3); __pyx_t_15++; if (unlikely(0 < 0)) __PYX_ERR(0, 990, __pyx_L1_error)
+          #else
+          __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_15); __pyx_t_15++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 990, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          #endif
+        } else {
+          if (__pyx_t_15 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_15); __Pyx_INCREF(__pyx_t_3); __pyx_t_15++; if (unlikely(0 < 0)) __PYX_ERR(0, 990, __pyx_L1_error)
+          #else
+          __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_15); __pyx_t_15++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 990, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          #endif
+        }
+      } else {
+        __pyx_t_3 = __pyx_t_8(__pyx_t_4);
+        if (unlikely(!__pyx_t_3)) {
+          PyObject* exc_type = PyErr_Occurred();
+          if (exc_type) {
+            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+            else __PYX_ERR(0, 990, __pyx_L1_error)
+          }
+          break;
+        }
+        __Pyx_GOTREF(__pyx_t_3);
+      }
+      __Pyx_XDECREF_SET(__pyx_v_value, __pyx_t_3);
+      __pyx_t_3 = 0;
+/* … */
+      __pyx_L22_continue:;
+    }
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+0991:                 X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, col, value)
+
      __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 991, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_6);
+      __pyx_t_12 = NULL;
+      __pyx_t_14 = 0;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
+        __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_6);
+        if (likely(__pyx_t_12)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+          __Pyx_INCREF(__pyx_t_12);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_6, function);
+          __pyx_t_14 = 1;
+        }
+      }
+      #if CYTHON_FAST_PYCALL
+      if (PyFunction_Check(__pyx_t_6)) {
+        PyObject *__pyx_temp[6] = {__pyx_t_12, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_col, __pyx_v_value};
+        __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 5+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __Pyx_GOTREF(__pyx_t_3);
+      } else
+      #endif
+      #if CYTHON_FAST_PYCCALL
+      if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
+        PyObject *__pyx_temp[6] = {__pyx_t_12, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_col, __pyx_v_value};
+        __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 5+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __Pyx_GOTREF(__pyx_t_3);
+      } else
+      #endif
+      {
+        __pyx_t_5 = PyTuple_New(5+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 991, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        if (__pyx_t_12) {
+          __Pyx_GIVEREF(__pyx_t_12); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_12); __pyx_t_12 = NULL;
+        }
+        __Pyx_INCREF(__pyx_v_X);
+        __Pyx_GIVEREF(__pyx_v_X);
+        PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_14, __pyx_v_X);
+        __Pyx_INCREF(__pyx_v_treatment_idx);
+        __Pyx_GIVEREF(__pyx_v_treatment_idx);
+        PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_14, __pyx_v_treatment_idx);
+        __Pyx_INCREF(__pyx_v_y);
+        __Pyx_GIVEREF(__pyx_v_y);
+        PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_14, __pyx_v_y);
+        __Pyx_INCREF(__pyx_v_col);
+        __Pyx_GIVEREF(__pyx_v_col);
+        PyTuple_SET_ITEM(__pyx_t_5, 3+__pyx_t_14, __pyx_v_col);
+        __Pyx_INCREF(__pyx_v_value);
+        __Pyx_GIVEREF(__pyx_v_value);
+        PyTuple_SET_ITEM(__pyx_t_5, 4+__pyx_t_14, __pyx_v_value);
+        __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_5, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+      if ((likely(PyTuple_CheckExact(__pyx_t_3))) || (PyList_CheckExact(__pyx_t_3))) {
+        PyObject* sequence = __pyx_t_3;
+        Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+        if (unlikely(size != 6)) {
+          if (size > 6) __Pyx_RaiseTooManyValuesError(6);
+          else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+          __PYX_ERR(0, 991, __pyx_L1_error)
+        }
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        if (likely(PyTuple_CheckExact(sequence))) {
+          __pyx_t_6 = PyTuple_GET_ITEM(sequence, 0); 
+          __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
+          __pyx_t_12 = PyTuple_GET_ITEM(sequence, 2); 
+          __pyx_t_11 = PyTuple_GET_ITEM(sequence, 3); 
+          __pyx_t_9 = PyTuple_GET_ITEM(sequence, 4); 
+          __pyx_t_16 = PyTuple_GET_ITEM(sequence, 5); 
+        } else {
+          __pyx_t_6 = PyList_GET_ITEM(sequence, 0); 
+          __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
+          __pyx_t_12 = PyList_GET_ITEM(sequence, 2); 
+          __pyx_t_11 = PyList_GET_ITEM(sequence, 3); 
+          __pyx_t_9 = PyList_GET_ITEM(sequence, 4); 
+          __pyx_t_16 = PyList_GET_ITEM(sequence, 5); 
+        }
+        __Pyx_INCREF(__pyx_t_6);
+        __Pyx_INCREF(__pyx_t_5);
+        __Pyx_INCREF(__pyx_t_12);
+        __Pyx_INCREF(__pyx_t_11);
+        __Pyx_INCREF(__pyx_t_9);
+        __Pyx_INCREF(__pyx_t_16);
+        #else
+        {
+          Py_ssize_t i;
+          PyObject** temps[6] = {&__pyx_t_6,&__pyx_t_5,&__pyx_t_12,&__pyx_t_11,&__pyx_t_9,&__pyx_t_16};
+          for (i=0; i < 6; i++) {
+            PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 991, __pyx_L1_error)
+            __Pyx_GOTREF(item);
+            *(temps[i]) = item;
+          }
+        }
+        #endif
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      } else {
+        Py_ssize_t index = -1;
+        PyObject** temps[6] = {&__pyx_t_6,&__pyx_t_5,&__pyx_t_12,&__pyx_t_11,&__pyx_t_9,&__pyx_t_16};
+        __pyx_t_17 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_17)) __PYX_ERR(0, 991, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_17);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_7 = Py_TYPE(__pyx_t_17)->tp_iternext;
+        for (index=0; index < 6; index++) {
+          PyObject* item = __pyx_t_7(__pyx_t_17); if (unlikely(!item)) goto __pyx_L24_unpacking_failed;
+          __Pyx_GOTREF(item);
+          *(temps[index]) = item;
+        }
+        if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_17), 6) < 0) __PYX_ERR(0, 991, __pyx_L1_error)
+        __pyx_t_7 = NULL;
+        __Pyx_DECREF(__pyx_t_17); __pyx_t_17 = 0;
+        goto __pyx_L25_unpacking_done;
+        __pyx_L24_unpacking_failed:;
+        __Pyx_DECREF(__pyx_t_17); __pyx_t_17 = 0;
+        __pyx_t_7 = NULL;
+        if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+        __PYX_ERR(0, 991, __pyx_L1_error)
+        __pyx_L25_unpacking_done:;
+      }
+      __Pyx_XDECREF_SET(__pyx_v_X_l, __pyx_t_6);
+      __pyx_t_6 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_X_r, __pyx_t_5);
+      __pyx_t_5 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_w_l, __pyx_t_12);
+      __pyx_t_12 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_w_r, __pyx_t_11);
+      __pyx_t_11 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_y_l, __pyx_t_9);
+      __pyx_t_9 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_y_r, __pyx_t_16);
+      __pyx_t_16 = 0;
+
 0992:                 # check the split validity on min_samples_leaf  372
+
+0993:                 if (len(X_l) < min_samples_leaf or len(X_r) < min_samples_leaf):
+
      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 993, __pyx_L1_error)
+      __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_16 = PyObject_RichCompare(__pyx_t_3, __pyx_v_min_samples_leaf, Py_LT); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      if (!__pyx_t_13) {
+      } else {
+        __pyx_t_2 = __pyx_t_13;
+        goto __pyx_L27_bool_binop_done;
+      }
+      __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 993, __pyx_L1_error)
+      __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __pyx_t_3 = PyObject_RichCompare(__pyx_t_16, __pyx_v_min_samples_leaf, Py_LT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 993, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_2 = __pyx_t_13;
+      __pyx_L27_bool_binop_done:;
+      if (__pyx_t_2) {
+/* … */
+      }
+
+0994:                     continue
+
        goto __pyx_L22_continue;
+
 0995:                 # summarize notes
+
 0996:                 # Gain -- Entropy or Gini
+
+0997:                 p = float(len(X_l)) / len(X)
+
      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 997, __pyx_L1_error)
+      __pyx_t_19 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_19 == ((Py_ssize_t)-1))) __PYX_ERR(0, 997, __pyx_L1_error)
+      __pyx_v_p = (((double)__pyx_t_18) / __pyx_t_19);
+
+0998:                 leftNodeSummary = self.tree_node_summary(w_l, y_l,
+
      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 998, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_16 = PyTuple_New(2); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 998, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __Pyx_INCREF(__pyx_v_w_l);
+      __Pyx_GIVEREF(__pyx_v_w_l);
+      PyTuple_SET_ITEM(__pyx_t_16, 0, __pyx_v_w_l);
+      __Pyx_INCREF(__pyx_v_y_l);
+      __Pyx_GIVEREF(__pyx_v_y_l);
+      PyTuple_SET_ITEM(__pyx_t_16, 1, __pyx_v_y_l);
+/* … */
+      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 998, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_leftNodeSummary, __pyx_t_11);
+      __pyx_t_11 = 0;
+
+0999:                                                          min_samples_treatment=min_samples_treatment,
+
      __pyx_t_9 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 999, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
+
+1000:                                                          n_reg=n_reg,
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
+
+1001:                                                          parentNodeSummary=currentNodeSummary)
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
+
 1002:
+
+1003:                 rightNodeSummary = self.tree_node_summary(w_r, y_r,
+
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1003, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_11);
+      __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1003, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __Pyx_INCREF(__pyx_v_w_r);
+      __Pyx_GIVEREF(__pyx_v_w_r);
+      PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_v_w_r);
+      __Pyx_INCREF(__pyx_v_y_r);
+      __Pyx_GIVEREF(__pyx_v_y_r);
+      PyTuple_SET_ITEM(__pyx_t_9, 1, __pyx_v_y_r);
+/* … */
+      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1003, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_rightNodeSummary, __pyx_t_3);
+      __pyx_t_3 = 0;
+
+1004:                                                           min_samples_treatment=min_samples_treatment,
+
      __pyx_t_16 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1004, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
+
+1005:                                                           n_reg=n_reg,
+
      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
+
+1006:                                                           parentNodeSummary=currentNodeSummary)
+
      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
+
 1007:
+
 1008:                 # check the split validity on min_samples_treatment
+
+1009:                 assert len(leftNodeSummary) == len(rightNodeSummary)
+
      #ifndef CYTHON_WITHOUT_ASSERTIONS
+      if (unlikely(!Py_OptimizeFlag)) {
+        __pyx_t_19 = PyObject_Length(__pyx_v_leftNodeSummary); if (unlikely(__pyx_t_19 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1009, __pyx_L1_error)
+        __pyx_t_18 = PyObject_Length(__pyx_v_rightNodeSummary); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1009, __pyx_L1_error)
+        if (unlikely(!((__pyx_t_19 == __pyx_t_18) != 0))) {
+          PyErr_SetNone(PyExc_AssertionError);
+          __PYX_ERR(0, 1009, __pyx_L1_error)
+        }
+      }
+      #endif
+
 1010:
+
+1011:                 node_mst = min([stat[1] for stat in leftNodeSummary + rightNodeSummary])
+
      __pyx_t_3 = PyList_New(0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1011, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_16 = PyNumber_Add(__pyx_v_leftNodeSummary, __pyx_v_rightNodeSummary); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      if (likely(PyList_CheckExact(__pyx_t_16)) || PyTuple_CheckExact(__pyx_t_16)) {
+        __pyx_t_9 = __pyx_t_16; __Pyx_INCREF(__pyx_t_9); __pyx_t_18 = 0;
+        __pyx_t_20 = NULL;
+      } else {
+        __pyx_t_18 = -1; __pyx_t_9 = PyObject_GetIter(__pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1011, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_20 = Py_TYPE(__pyx_t_9)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1011, __pyx_L1_error)
+      }
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      for (;;) {
+        if (likely(!__pyx_t_20)) {
+          if (likely(PyList_CheckExact(__pyx_t_9))) {
+            if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_9)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_16 = PyList_GET_ITEM(__pyx_t_9, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1011, __pyx_L1_error)
+            #else
+            __pyx_t_16 = PySequence_ITEM(__pyx_t_9, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_16);
+            #endif
+          } else {
+            if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_9)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_16 = PyTuple_GET_ITEM(__pyx_t_9, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1011, __pyx_L1_error)
+            #else
+            __pyx_t_16 = PySequence_ITEM(__pyx_t_9, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_16);
+            #endif
+          }
+        } else {
+          __pyx_t_16 = __pyx_t_20(__pyx_t_9);
+          if (unlikely(!__pyx_t_16)) {
+            PyObject* exc_type = PyErr_Occurred();
+            if (exc_type) {
+              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+              else __PYX_ERR(0, 1011, __pyx_L1_error)
+            }
+            break;
+          }
+          __Pyx_GOTREF(__pyx_t_16);
+        }
+        __Pyx_XDECREF_SET(__pyx_v_stat, __pyx_t_16);
+        __pyx_t_16 = 0;
+        __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_stat, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        if (unlikely(__Pyx_ListComp_Append(__pyx_t_3, (PyObject*)__pyx_t_16))) __PYX_ERR(0, 1011, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __pyx_t_9 = __Pyx_PyObject_CallOneArg(__pyx_builtin_min, __pyx_t_3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1011, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __Pyx_XDECREF_SET(__pyx_v_node_mst, __pyx_t_9);
+      __pyx_t_9 = 0;
+
+1012:                 if node_mst < min_samples_treatment:
+
      __pyx_t_9 = PyObject_RichCompare(__pyx_v_node_mst, __pyx_v_min_samples_treatment, Py_LT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1012, __pyx_L1_error)
+      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1012, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      if (__pyx_t_2) {
+/* … */
+      }
+
+1013:                     continue
+
        goto __pyx_L22_continue;
+
 1014:
+
 1015:                 # evaluate the split
+
+1016:                 if self.evaluationFunction == self.evaluate_CTS:
+
      __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1016, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1016, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      __pyx_t_16 = PyObject_RichCompare(__pyx_t_9, __pyx_t_3, Py_EQ); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1016, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1016, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      if (__pyx_t_2) {
+/* … */
+        goto __pyx_L32;
+      }
+
+1017:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
+
        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1017, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_9 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
+          __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_3);
+          if (likely(__pyx_t_9)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+            __Pyx_INCREF(__pyx_t_9);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_3, function);
+          }
+        }
+        __pyx_t_16 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_9, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_leftNodeSummary);
+        __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+        if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1017, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_16);
+        __pyx_t_16 = 0;
+
+1018:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
+
        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1018, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_9 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
+          __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_3);
+          if (likely(__pyx_t_9)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+            __Pyx_INCREF(__pyx_t_9);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_3, function);
+          }
+        }
+        __pyx_t_16 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_9, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_rightNodeSummary);
+        __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+        if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1018, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_16);
+        __pyx_t_16 = 0;
+
+1019:                     gain = (currentScore - p * leftScore1 - (1 - p) * rightScore2)
+
        __pyx_t_16 = PyFloat_FromDouble(__pyx_v_p); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_3 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_16 = PyNumber_Subtract(__pyx_v_currentScore, __pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = PyFloat_FromDouble((1.0 - __pyx_v_p)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = PyNumber_Subtract(__pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1020:                     gain_for_imp = (len(X) * currentScore - len(X_l) * leftScore1 - len(X_r) * rightScore2)
+
        __pyx_t_18 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_currentScore); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_16 = PyNumber_Multiply(__pyx_t_3, __pyx_v_leftScore1); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = PyNumber_Subtract(__pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_16 = PyNumber_Subtract(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_16);
+        __pyx_t_16 = 0;
+
+1021:                 elif self.evaluationFunction == self.evaluate_DDP:
+
      __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1021, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1021, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __pyx_t_3 = PyObject_RichCompare(__pyx_t_16, __pyx_t_9, Py_EQ); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1021, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1021, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      if (__pyx_t_2) {
+/* … */
+        goto __pyx_L32;
+      }
+
+1022:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
+
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1022, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_16 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
+          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
+          if (likely(__pyx_t_16)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
+            __Pyx_INCREF(__pyx_t_16);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_9, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_leftNodeSummary);
+        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1022, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1023:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
+
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1023, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_16 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
+          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
+          if (likely(__pyx_t_16)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
+            __Pyx_INCREF(__pyx_t_16);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_9, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_rightNodeSummary);
+        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1023, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1024:                     gain = np.abs(leftScore1 - rightScore2)
+
        __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1024, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_abs); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1024, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __pyx_t_9 = PyNumber_Subtract(__pyx_v_leftScore1, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1024, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_11 = NULL;
+        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_16))) {
+          __pyx_t_11 = PyMethod_GET_SELF(__pyx_t_16);
+          if (likely(__pyx_t_11)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_16);
+            __Pyx_INCREF(__pyx_t_11);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_16, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_11) ? __Pyx_PyObject_Call2Args(__pyx_t_16, __pyx_t_11, __pyx_t_9) : __Pyx_PyObject_CallOneArg(__pyx_t_16, __pyx_t_9);
+        __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1024, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1025:                     gain_for_imp = np.abs(len(X_l) * leftScore1 - len(X_r) * rightScore2)
+
        __Pyx_GetModuleGlobalName(__pyx_t_16, __pyx_n_s_np); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_t_16, __pyx_n_s_abs); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_11 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_11);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_12 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_16 = PyNumber_Subtract(__pyx_t_11, __pyx_t_12); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
+        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __pyx_t_12 = NULL;
+        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_9))) {
+          __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_9);
+          if (likely(__pyx_t_12)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
+            __Pyx_INCREF(__pyx_t_12);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_9, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_12) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_12, __pyx_t_16) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_t_16);
+        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1025, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_3);
+        __pyx_t_3 = 0;
+
 1026:                 else:
+
+1027:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
+
      /*else*/ {
+        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1027, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_16 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
+          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
+          if (likely(__pyx_t_16)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
+            __Pyx_INCREF(__pyx_t_16);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_9, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_leftNodeSummary);
+        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1027, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1028:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
+
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1028, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_16 = NULL;
+        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
+          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
+          if (likely(__pyx_t_16)) {
+            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
+            __Pyx_INCREF(__pyx_t_16);
+            __Pyx_INCREF(function);
+            __Pyx_DECREF_SET(__pyx_t_9, function);
+          }
+        }
+        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_rightNodeSummary);
+        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
+        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1028, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_3);
+        __pyx_t_3 = 0;
+
+1029:                     gain = (p * leftScore1 + (1 - p) * rightScore2 - currentScore)
+
        __pyx_t_3 = PyFloat_FromDouble(__pyx_v_p); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_leftScore1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = PyFloat_FromDouble((1.0 - __pyx_v_p)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __pyx_t_16 = PyNumber_Multiply(__pyx_t_3, __pyx_v_rightScore2); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __pyx_t_3 = PyNumber_Add(__pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_16 = PyNumber_Subtract(__pyx_t_3, __pyx_v_currentScore); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1029, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_16);
+        __pyx_t_16 = 0;
+
+1030:                     gain_for_imp = (len(X_l) * leftScore1 + len(X_r) * rightScore2 - len(X) * currentScore)
+
        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_3 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __pyx_t_16 = PyNumber_Add(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __pyx_t_18 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __pyx_t_9 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_3 = PyNumber_Multiply(__pyx_t_9, __pyx_v_currentScore); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __pyx_t_9 = PyNumber_Subtract(__pyx_t_16, __pyx_t_3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_9);
+        __pyx_t_9 = 0;
+
+1031:                     if self.normalization:
+
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_normalization); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1031, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1031, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        if (__pyx_t_2) {
+/* … */
+          goto __pyx_L33;
+        }
+
+1032:                         n_c = currentNodeSummary[0][1]
+
          __pyx_t_9 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1032, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_9, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1032, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+          __Pyx_DECREF_SET(__pyx_v_n_c, __pyx_t_3);
+          __pyx_t_3 = 0;
+
+1033:                         n_c_left = leftNodeSummary[0][1]
+
          __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_leftNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1033, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          __pyx_t_9 = __Pyx_GetItemInt(__pyx_t_3, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1033, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+          __Pyx_XDECREF_SET(__pyx_v_n_c_left, __pyx_t_9);
+          __pyx_t_9 = 0;
+
+1034:                         n_t = [tr[1] for tr in currentNodeSummary[1:]]
+
          __pyx_t_9 = PyList_New(0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1034, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __pyx_t_3 = __Pyx_PyObject_GetSlice(__pyx_v_currentNodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
+            __pyx_t_16 = __pyx_t_3; __Pyx_INCREF(__pyx_t_16); __pyx_t_18 = 0;
+            __pyx_t_20 = NULL;
+          } else {
+            __pyx_t_18 = -1; __pyx_t_16 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1034, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_16);
+            __pyx_t_20 = Py_TYPE(__pyx_t_16)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1034, __pyx_L1_error)
+          }
+          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+          for (;;) {
+            if (likely(!__pyx_t_20)) {
+              if (likely(PyList_CheckExact(__pyx_t_16))) {
+                if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_16)) break;
+                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+                __pyx_t_3 = PyList_GET_ITEM(__pyx_t_16, __pyx_t_18); __Pyx_INCREF(__pyx_t_3); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1034, __pyx_L1_error)
+                #else
+                __pyx_t_3 = PySequence_ITEM(__pyx_t_16, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
+                __Pyx_GOTREF(__pyx_t_3);
+                #endif
+              } else {
+                if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_16)) break;
+                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+                __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_16, __pyx_t_18); __Pyx_INCREF(__pyx_t_3); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1034, __pyx_L1_error)
+                #else
+                __pyx_t_3 = PySequence_ITEM(__pyx_t_16, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
+                __Pyx_GOTREF(__pyx_t_3);
+                #endif
+              }
+            } else {
+              __pyx_t_3 = __pyx_t_20(__pyx_t_16);
+              if (unlikely(!__pyx_t_3)) {
+                PyObject* exc_type = PyErr_Occurred();
+                if (exc_type) {
+                  if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+                  else __PYX_ERR(0, 1034, __pyx_L1_error)
+                }
+                break;
+              }
+              __Pyx_GOTREF(__pyx_t_3);
+            }
+            __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_3);
+            __pyx_t_3 = 0;
+            __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_tr, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_3);
+            if (unlikely(__Pyx_ListComp_Append(__pyx_t_9, (PyObject*)__pyx_t_3))) __PYX_ERR(0, 1034, __pyx_L1_error)
+            __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+          }
+          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+          __Pyx_DECREF_SET(__pyx_v_n_t, __pyx_t_9);
+          __pyx_t_9 = 0;
+
+1035:                         n_t_left = [tr[1] for tr in leftNodeSummary[1:]]
+
          __pyx_t_9 = PyList_New(0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1035, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __pyx_t_16 = __Pyx_PyObject_GetSlice(__pyx_v_leftNodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_16);
+          if (likely(PyList_CheckExact(__pyx_t_16)) || PyTuple_CheckExact(__pyx_t_16)) {
+            __pyx_t_3 = __pyx_t_16; __Pyx_INCREF(__pyx_t_3); __pyx_t_18 = 0;
+            __pyx_t_20 = NULL;
+          } else {
+            __pyx_t_18 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1035, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_3);
+            __pyx_t_20 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1035, __pyx_L1_error)
+          }
+          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+          for (;;) {
+            if (likely(!__pyx_t_20)) {
+              if (likely(PyList_CheckExact(__pyx_t_3))) {
+                if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_3)) break;
+                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+                __pyx_t_16 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1035, __pyx_L1_error)
+                #else
+                __pyx_t_16 = PySequence_ITEM(__pyx_t_3, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
+                __Pyx_GOTREF(__pyx_t_16);
+                #endif
+              } else {
+                if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
+                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+                __pyx_t_16 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1035, __pyx_L1_error)
+                #else
+                __pyx_t_16 = PySequence_ITEM(__pyx_t_3, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
+                __Pyx_GOTREF(__pyx_t_16);
+                #endif
+              }
+            } else {
+              __pyx_t_16 = __pyx_t_20(__pyx_t_3);
+              if (unlikely(!__pyx_t_16)) {
+                PyObject* exc_type = PyErr_Occurred();
+                if (exc_type) {
+                  if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+                  else __PYX_ERR(0, 1035, __pyx_L1_error)
+                }
+                break;
+              }
+              __Pyx_GOTREF(__pyx_t_16);
+            }
+            __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_16);
+            __pyx_t_16 = 0;
+            __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_tr, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_16);
+            if (unlikely(__Pyx_ListComp_Append(__pyx_t_9, (PyObject*)__pyx_t_16))) __PYX_ERR(0, 1035, __pyx_L1_error)
+            __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+          }
+          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+          __Pyx_XDECREF_SET(__pyx_v_n_t_left, ((PyObject*)__pyx_t_9));
+          __pyx_t_9 = 0;
+
 1036:
+
+1037:                         norm_factor = self.normI(n_c, n_c_left, n_t, n_t_left, alpha=0.9)
+
          __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_normI); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1037, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __pyx_t_3 = PyTuple_New(4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1037, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_3);
+          __Pyx_INCREF(__pyx_v_n_c);
+          __Pyx_GIVEREF(__pyx_v_n_c);
+          PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_n_c);
+          __Pyx_INCREF(__pyx_v_n_c_left);
+          __Pyx_GIVEREF(__pyx_v_n_c_left);
+          PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_v_n_c_left);
+          __Pyx_INCREF(__pyx_v_n_t);
+          __Pyx_GIVEREF(__pyx_v_n_t);
+          PyTuple_SET_ITEM(__pyx_t_3, 2, __pyx_v_n_t);
+          __Pyx_INCREF(__pyx_v_n_t_left);
+          __Pyx_GIVEREF(__pyx_v_n_t_left);
+          PyTuple_SET_ITEM(__pyx_t_3, 3, __pyx_v_n_t_left);
+          __pyx_t_16 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1037, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_16);
+          if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_alpha, __pyx_float_0_9) < 0) __PYX_ERR(0, 1037, __pyx_L1_error)
+          __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_9, __pyx_t_3, __pyx_t_16); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1037, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_12);
+          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+          __Pyx_XDECREF_SET(__pyx_v_norm_factor, __pyx_t_12);
+          __pyx_t_12 = 0;
+
 1038:                     else:
+
+1039:                         norm_factor = 1
+
        /*else*/ {
+          __Pyx_INCREF(__pyx_int_1);
+          __Pyx_XDECREF_SET(__pyx_v_norm_factor, __pyx_int_1);
+        }
+        __pyx_L33:;
+
+1040:                     gain = gain / norm_factor
+
        __pyx_t_12 = __Pyx_PyNumber_Divide(__pyx_v_gain, __pyx_v_norm_factor); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1040, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_DECREF_SET(__pyx_v_gain, __pyx_t_12);
+        __pyx_t_12 = 0;
+      }
+      __pyx_L32:;
+
+1041:                 if (gain > bestGain and len(X_l) > min_samples_leaf and len(X_r) > min_samples_leaf):
+
      __pyx_t_12 = PyObject_RichCompare(__pyx_v_gain, __pyx_v_bestGain, Py_GT); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      if (__pyx_t_13) {
+      } else {
+        __pyx_t_2 = __pyx_t_13;
+        goto __pyx_L39_bool_binop_done;
+      }
+      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __pyx_t_12 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+      __pyx_t_16 = PyObject_RichCompare(__pyx_t_12, __pyx_v_min_samples_leaf, Py_GT); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      if (__pyx_t_13) {
+      } else {
+        __pyx_t_2 = __pyx_t_13;
+        goto __pyx_L39_bool_binop_done;
+      }
+      __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __pyx_t_12 = PyObject_RichCompare(__pyx_t_16, __pyx_v_min_samples_leaf, Py_GT); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      __pyx_t_2 = __pyx_t_13;
+      __pyx_L39_bool_binop_done:;
+      if (__pyx_t_2) {
+/* … */
+      }
+
+1042:                     bestGain = gain
+
        __Pyx_INCREF(__pyx_v_gain);
+        __Pyx_DECREF_SET(__pyx_v_bestGain, __pyx_v_gain);
+
+1043:                     bestGainImp = gain_for_imp
+
        __Pyx_INCREF(__pyx_v_gain_for_imp);
+        __Pyx_DECREF_SET(__pyx_v_bestGainImp, __pyx_v_gain_for_imp);
+
+1044:                     bestAttribute = (col, value)
+
        __pyx_t_12 = PyTuple_New(2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1044, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_INCREF(__pyx_v_col);
+        __Pyx_GIVEREF(__pyx_v_col);
+        PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_v_col);
+        __Pyx_INCREF(__pyx_v_value);
+        __Pyx_GIVEREF(__pyx_v_value);
+        PyTuple_SET_ITEM(__pyx_t_12, 1, __pyx_v_value);
+        __Pyx_DECREF_SET(__pyx_v_bestAttribute, ((PyObject*)__pyx_t_12));
+        __pyx_t_12 = 0;
+
+1045:                     best_set_left = [X_l, w_l, y_l]
+
        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1045, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_INCREF(__pyx_v_X_l);
+        __Pyx_GIVEREF(__pyx_v_X_l);
+        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_v_X_l);
+        __Pyx_INCREF(__pyx_v_w_l);
+        __Pyx_GIVEREF(__pyx_v_w_l);
+        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_v_w_l);
+        __Pyx_INCREF(__pyx_v_y_l);
+        __Pyx_GIVEREF(__pyx_v_y_l);
+        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_v_y_l);
+        __Pyx_XDECREF_SET(__pyx_v_best_set_left, ((PyObject*)__pyx_t_12));
+        __pyx_t_12 = 0;
+
+1046:                     best_set_right = [X_r, w_r, y_r]
+
        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1046, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_INCREF(__pyx_v_X_r);
+        __Pyx_GIVEREF(__pyx_v_X_r);
+        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_v_X_r);
+        __Pyx_INCREF(__pyx_v_w_r);
+        __Pyx_GIVEREF(__pyx_v_w_r);
+        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_v_w_r);
+        __Pyx_INCREF(__pyx_v_y_r);
+        __Pyx_GIVEREF(__pyx_v_y_r);
+        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_v_y_r);
+        __Pyx_XDECREF_SET(__pyx_v_best_set_right, ((PyObject*)__pyx_t_12));
+        __pyx_t_12 = 0;
+
 1047:
+
+1048:         dcY = {'impurity': '%.3f' % currentScore, 'samples': '%d' % len(X)}
+
  __pyx_t_10 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __pyx_t_4 = __Pyx_PyString_FormatSafe(__pyx_kp_s_3f, __pyx_v_currentScore); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  if (PyDict_SetItem(__pyx_t_10, __pyx_n_s_impurity, __pyx_t_4) < 0) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __pyx_t_1 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_1 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_12 = __Pyx_PyString_Format(__pyx_kp_s_d, __pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (PyDict_SetItem(__pyx_t_10, __pyx_n_s_samples, __pyx_t_12) < 0) __PYX_ERR(0, 1048, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __pyx_v_dcY = ((PyObject*)__pyx_t_10);
+  __pyx_t_10 = 0;
+
 1049:         # Add treatment size
+
+1050:         dcY['group_size'] = ''
+
  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_group_size, __pyx_kp_s__4) < 0)) __PYX_ERR(0, 1050, __pyx_L1_error)
+
+1051:         for i, summary in enumerate(currentNodeSummary):
+
  __Pyx_INCREF(__pyx_int_0);
+  __pyx_t_10 = __pyx_int_0;
+  if (likely(PyList_CheckExact(__pyx_v_currentNodeSummary)) || PyTuple_CheckExact(__pyx_v_currentNodeSummary)) {
+    __pyx_t_12 = __pyx_v_currentNodeSummary; __Pyx_INCREF(__pyx_t_12); __pyx_t_1 = 0;
+    __pyx_t_8 = NULL;
+  } else {
+    __pyx_t_1 = -1; __pyx_t_12 = PyObject_GetIter(__pyx_v_currentNodeSummary); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1051, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __pyx_t_8 = Py_TYPE(__pyx_t_12)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1051, __pyx_L1_error)
+  }
+  for (;;) {
+    if (likely(!__pyx_t_8)) {
+      if (likely(PyList_CheckExact(__pyx_t_12))) {
+        if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_12)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_12, __pyx_t_1); __Pyx_INCREF(__pyx_t_4); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 1051, __pyx_L1_error)
+        #else
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_12, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        #endif
+      } else {
+        if (__pyx_t_1 >= PyTuple_GET_SIZE(__pyx_t_12)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_12, __pyx_t_1); __Pyx_INCREF(__pyx_t_4); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 1051, __pyx_L1_error)
+        #else
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_12, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        #endif
+      }
+    } else {
+      __pyx_t_4 = __pyx_t_8(__pyx_t_12);
+      if (unlikely(!__pyx_t_4)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1051, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_4);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_summary, __pyx_t_4);
+    __pyx_t_4 = 0;
+    __Pyx_INCREF(__pyx_t_10);
+    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_10);
+    __pyx_t_4 = __Pyx_PyInt_AddObjC(__pyx_t_10, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_10);
+    __pyx_t_10 = __pyx_t_4;
+    __pyx_t_4 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
+1052:             dcY['group_size'] += ' ' + self.classes_[i] + ': ' + str(summary[1])
+
    __Pyx_INCREF(__pyx_n_s_group_size);
+    __pyx_t_21 = __pyx_n_s_group_size;
+    __pyx_t_4 = __Pyx_PyDict_GetItem(__pyx_v_dcY, __pyx_t_21); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_t_16, __pyx_v_i); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __pyx_t_16 = PyNumber_Add(__pyx_kp_s__5, __pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __pyx_t_3 = PyNumber_Add(__pyx_t_16, __pyx_kp_s__6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_summary, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyString_Type)), __pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __pyx_t_16 = PyNumber_Add(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_4, __pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_t_21, __pyx_t_9) < 0)) __PYX_ERR(0, 1052, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __Pyx_DECREF(__pyx_t_21); __pyx_t_21 = 0;
+
+1053:         dcY['upliftScore'] = [round(upliftScore[0], 4), round(upliftScore[1], 4)]
+
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  __Pyx_INCREF(__pyx_int_4);
+  __Pyx_GIVEREF(__pyx_int_4);
+  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
+  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_12);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 1));
+  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 1));
+  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 1));
+  __Pyx_INCREF(__pyx_int_4);
+  __Pyx_GIVEREF(__pyx_int_4);
+  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
+  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  __pyx_t_10 = PyList_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_GIVEREF(__pyx_t_12);
+  PyList_SET_ITEM(__pyx_t_10, 0, __pyx_t_12);
+  __Pyx_GIVEREF(__pyx_t_9);
+  PyList_SET_ITEM(__pyx_t_10, 1, __pyx_t_9);
+  __pyx_t_12 = 0;
+  __pyx_t_9 = 0;
+  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_upliftScore, __pyx_t_10) < 0)) __PYX_ERR(0, 1053, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+
+1054:         dcY['matchScore'] = round(upliftScore[0], 4)
+
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1054, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_10);
+  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 0));
+  __Pyx_INCREF(__pyx_int_4);
+  __Pyx_GIVEREF(__pyx_int_4);
+  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
+  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1054, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_9);
+  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_matchScore, __pyx_t_9) < 0)) __PYX_ERR(0, 1054, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+
 1055:
+
+1056:         if bestGain > 0 and depth < max_depth:
+
  __pyx_t_9 = PyObject_RichCompare(__pyx_v_bestGain, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1056, __pyx_L1_error)
+  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  if (__pyx_t_13) {
+  } else {
+    __pyx_t_2 = __pyx_t_13;
+    goto __pyx_L45_bool_binop_done;
+  }
+  __pyx_t_9 = PyObject_RichCompare(__pyx_v_depth, __pyx_v_max_depth, Py_LT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1056, __pyx_L1_error)
+  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+  __pyx_t_2 = __pyx_t_13;
+  __pyx_L45_bool_binop_done:;
+  if (__pyx_t_2) {
+/* … */
+  }
+
+1057:             self.feature_imp_dict[bestAttribute[0]] += bestGainImp
+
    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1057, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 1057, __pyx_L1_error)
+    }
+    __Pyx_INCREF(PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0));
+    __pyx_t_10 = PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0);
+    __pyx_t_12 = __Pyx_PyObject_GetItem(__pyx_t_9, __pyx_t_10); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1057, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __pyx_t_16 = PyNumber_InPlaceAdd(__pyx_t_12, __pyx_v_bestGainImp); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1057, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    if (unlikely(PyObject_SetItem(__pyx_t_9, __pyx_t_10, __pyx_t_16) < 0)) __PYX_ERR(0, 1057, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+
+1058:             trueBranch = self.growDecisionTreeFrom(
+
    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1058, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+/* … */
+    __pyx_t_10 = PySequence_Tuple(__pyx_v_best_set_left); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1058, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+/* … */
+    __pyx_t_12 = PyTuple_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1058, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __Pyx_INCREF(__pyx_v_max_depth);
+    __Pyx_GIVEREF(__pyx_v_max_depth);
+    PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_v_max_depth);
+    __Pyx_INCREF(__pyx_v_min_samples_leaf);
+    __Pyx_GIVEREF(__pyx_v_min_samples_leaf);
+    PyTuple_SET_ITEM(__pyx_t_12, 1, __pyx_v_min_samples_leaf);
+    __Pyx_GIVEREF(__pyx_t_16);
+    PyTuple_SET_ITEM(__pyx_t_12, 2, __pyx_t_16);
+    __pyx_t_16 = 0;
+    __pyx_t_16 = PyNumber_Add(__pyx_t_10, __pyx_t_12); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1058, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+/* … */
+    __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_9, __pyx_t_16, __pyx_t_12); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1058, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __pyx_v_trueBranch = __pyx_t_10;
+    __pyx_t_10 = 0;
+
+1059:                 *best_set_left, max_depth, min_samples_leaf,
+
    if (unlikely(!__pyx_v_best_set_left)) { __Pyx_RaiseUnboundLocalError("best_set_left"); __PYX_ERR(0, 1059, __pyx_L1_error) }
+
+1060:                 depth + 1, min_samples_treatment=min_samples_treatment,
+
    __pyx_t_16 = __Pyx_PyInt_AddObjC(__pyx_v_depth, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1060, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+/* … */
+    __pyx_t_12 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1060, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
+
+1061:                 n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
+
 1062:             )
+
+1063:             falseBranch = self.growDecisionTreeFrom(
+
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1063, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+/* … */
+    __pyx_t_12 = PySequence_Tuple(__pyx_v_best_set_right); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1063, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+/* … */
+    __pyx_t_9 = PyTuple_New(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1063, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __Pyx_INCREF(__pyx_v_max_depth);
+    __Pyx_GIVEREF(__pyx_v_max_depth);
+    PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_v_max_depth);
+    __Pyx_INCREF(__pyx_v_min_samples_leaf);
+    __Pyx_GIVEREF(__pyx_v_min_samples_leaf);
+    PyTuple_SET_ITEM(__pyx_t_9, 1, __pyx_v_min_samples_leaf);
+    __Pyx_GIVEREF(__pyx_t_16);
+    PyTuple_SET_ITEM(__pyx_t_9, 2, __pyx_t_16);
+    __pyx_t_16 = 0;
+    __pyx_t_16 = PyNumber_Add(__pyx_t_12, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1063, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+/* … */
+    __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1063, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_v_falseBranch = __pyx_t_12;
+    __pyx_t_12 = 0;
+
+1064:                 *best_set_right, max_depth, min_samples_leaf,
+
    if (unlikely(!__pyx_v_best_set_right)) { __Pyx_RaiseUnboundLocalError("best_set_right"); __PYX_ERR(0, 1064, __pyx_L1_error) }
+
+1065:                 depth + 1, min_samples_treatment=min_samples_treatment,
+
    __pyx_t_16 = __Pyx_PyInt_AddObjC(__pyx_v_depth, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1065, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+/* … */
+    __pyx_t_9 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1065, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
+
+1066:                 n_reg=n_reg, parentNodeSummary=currentNodeSummary
+
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
+
 1067:             )
+
 1068:
+
+1069:             return DecisionTree(
+
    __Pyx_XDECREF(__pyx_r);
+    __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1069, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_12);
+/* … */
+    __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1069, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_r = __pyx_t_16;
+    __pyx_t_16 = 0;
+    goto __pyx_L0;
+
+1070:                 classes_=self.classes_,
+
    __pyx_t_9 = __Pyx_PyDict_NewPresized(12); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1070, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1070, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_16) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+
+1071:                 col=bestAttribute[0], value=bestAttribute[1],
+
    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 1071, __pyx_L1_error)
+    }
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_col, PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0)) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
+      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
+      __PYX_ERR(0, 1071, __pyx_L1_error)
+    }
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_value, PyTuple_GET_ITEM(__pyx_v_bestAttribute, 1)) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+
+1072:                 trueBranch=trueBranch, falseBranch=falseBranch, summary=dcY,
+
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_trueBranch, __pyx_v_trueBranch) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_falseBranch, __pyx_v_falseBranch) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+
+1073:                 maxDiffTreatment=maxDiffTreatment, maxDiffSign=maxDiffSign,
+
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+
+1074:                 nodeSummary=currentNodeSummary,
+
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+
+1075:                 backupResults=self.uplift_classification_results(treatment_idx, y),
+
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1075, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_10);
+    __pyx_t_4 = NULL;
+    __pyx_t_14 = 0;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+      __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_10);
+      if (likely(__pyx_t_4)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+        __Pyx_INCREF(__pyx_t_4);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_10, function);
+        __pyx_t_14 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_10)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
+      __pyx_t_16 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_GOTREF(__pyx_t_16);
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+      PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
+      __pyx_t_16 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_GOTREF(__pyx_t_16);
+    } else
+    #endif
+    {
+      __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1075, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_3);
+      if (__pyx_t_4) {
+        __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_4); __pyx_t_4 = NULL;
+      }
+      __Pyx_INCREF(__pyx_v_treatment_idx);
+      __Pyx_GIVEREF(__pyx_v_treatment_idx);
+      PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_treatment_idx);
+      __Pyx_INCREF(__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_v_y);
+      __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, NULL); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_backupResults, __pyx_t_16) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+
+1076:                 bestTreatment=bestTreatment, upliftScore=upliftScore
+
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
+
 1077:             )
+
 1078:         else:
+
+1079:             if self.evaluationFunction == self.evaluate_CTS:
+
  /*else*/ {
+    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1079, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_16);
+    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1079, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __pyx_t_12 = PyObject_RichCompare(__pyx_t_16, __pyx_t_9, Py_EQ); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1079, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1079, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+    if (__pyx_t_2) {
+/* … */
+    }
+
+1080:                 return DecisionTree(
+
      __Pyx_XDECREF(__pyx_r);
+      __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1080, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+/* … */
+      __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1080, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __pyx_r = __pyx_t_16;
+      __pyx_t_16 = 0;
+      goto __pyx_L0;
+
+1081:                     classes_=self.classes_,
+
      __pyx_t_9 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1081, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1081, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_16) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+
+1082:                     results=self.uplift_classification_results(treatment_idx, y),
+
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1082, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __pyx_t_3 = NULL;
+      __pyx_t_14 = 0;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+        __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_10);
+        if (likely(__pyx_t_3)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+          __Pyx_INCREF(__pyx_t_3);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_10, function);
+          __pyx_t_14 = 1;
+        }
+      }
+      #if CYTHON_FAST_PYCALL
+      if (PyFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+        __pyx_t_16 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_GOTREF(__pyx_t_16);
+      } else
+      #endif
+      #if CYTHON_FAST_PYCCALL
+      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
+        __pyx_t_16 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
+        __Pyx_GOTREF(__pyx_t_16);
+      } else
+      #endif
+      {
+        __pyx_t_4 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1082, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        if (__pyx_t_3) {
+          __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3); __pyx_t_3 = NULL;
+        }
+        __Pyx_INCREF(__pyx_v_treatment_idx);
+        __Pyx_GIVEREF(__pyx_v_treatment_idx);
+        PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_14, __pyx_v_treatment_idx);
+        __Pyx_INCREF(__pyx_v_y);
+        __Pyx_GIVEREF(__pyx_v_y);
+        PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_14, __pyx_v_y);
+        __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_4, NULL); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_16);
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_results, __pyx_t_16) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+
+1083:                     summary=dcY, nodeSummary=currentNodeSummary,
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+
+1084:                     bestTreatment=bestTreatment, upliftScore=upliftScore
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
+
 1085:                 )
+
 1086:             else:
+
+1087:                 return DecisionTree(
+
    /*else*/ {
+      __Pyx_XDECREF(__pyx_r);
+      __Pyx_GetModuleGlobalName(__pyx_t_16, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1087, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_16);
+/* … */
+      __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_16, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1087, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
+      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+      __pyx_r = __pyx_t_12;
+      __pyx_t_12 = 0;
+      goto __pyx_L0;
+    }
+  }
+
+1088:                     classes_=self.classes_,
+
      __pyx_t_9 = __Pyx_PyDict_NewPresized(8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1088, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_9);
+      __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1088, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_12);
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_12) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+
+1089:                     results=self.uplift_classification_results(treatment_idx, y),
+
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1089, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_10);
+      __pyx_t_4 = NULL;
+      __pyx_t_14 = 0;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
+        __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_10);
+        if (likely(__pyx_t_4)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
+          __Pyx_INCREF(__pyx_t_4);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_10, function);
+          __pyx_t_14 = 1;
+        }
+      }
+      #if CYTHON_FAST_PYCALL
+      if (PyFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
+        __pyx_t_12 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_GOTREF(__pyx_t_12);
+      } else
+      #endif
+      #if CYTHON_FAST_PYCCALL
+      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
+        PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
+        __pyx_t_12 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
+        __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __Pyx_GOTREF(__pyx_t_12);
+      } else
+      #endif
+      {
+        __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1089, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        if (__pyx_t_4) {
+          __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_4); __pyx_t_4 = NULL;
+        }
+        __Pyx_INCREF(__pyx_v_treatment_idx);
+        __Pyx_GIVEREF(__pyx_v_treatment_idx);
+        PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_treatment_idx);
+        __Pyx_INCREF(__pyx_v_y);
+        __Pyx_GIVEREF(__pyx_v_y);
+        PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_v_y);
+        __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_12);
+        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+      }
+      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_results, __pyx_t_12) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
+
+1090:                     summary=dcY, maxDiffTreatment=maxDiffTreatment,
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+
+1091:                     maxDiffSign=maxDiffSign, nodeSummary=currentNodeSummary,
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+
+1092:                     bestTreatment=bestTreatment, upliftScore=upliftScore
+
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
+
 1093:                 )
+
 1094:
+
+1095:     @staticmethod
+
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1095, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_classify, __pyx_t_2) < 0) __PYX_ERR(0, 1096, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1096:     def classify(observations, tree, dataMissing=False):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify[] = "\n        Classifies (prediction) the observations according to the tree.\n\n        Args\n        ----\n        observations : list of list\n            The internal data format for the training data (combining X, Y, treatment).\n\n        dataMissing: boolean, optional (default = False)\n            An indicator for if data are missing or not.\n\n        Returns\n        -------\n        tree.results, tree.upliftScore :\n            The results in the leaf node.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify = {"classify", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_observations = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_v_dataMissing = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classify (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,&__pyx_n_s_dataMissing,0};
+    PyObject* values[3] = {0,0,0};
+    values[2] = ((PyObject *)((PyObject *)Py_False));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("classify", 0, 2, 3, 1); __PYX_ERR(0, 1096, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_dataMissing);
+          if (value) { values[2] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classify") < 0)) __PYX_ERR(0, 1096, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_observations = values[0];
+    __pyx_v_tree = values[1];
+    __pyx_v_dataMissing = values[2];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("classify", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1096, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify(__pyx_self, __pyx_v_observations, __pyx_v_tree, __pyx_v_dataMissing);
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree, PyObject *__pyx_v_dataMissing) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classify", 0);
+  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify, __pyx_empty_tuple, NULL);
+  if (unlikely(!__pyx_cur_scope)) {
+    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *)Py_None);
+    __Pyx_INCREF(Py_None);
+    __PYX_ERR(0, 1096, __pyx_L1_error)
+  } else {
+    __Pyx_GOTREF(__pyx_cur_scope);
+  }
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__61 = PyTuple_Pack(7, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_dataMissing, __pyx_n_s_classifyWithoutMissingData, __pyx_n_s_classifyWithoutMissingData, __pyx_n_s_classifyWithMissingData, __pyx_n_s_classifyWithMissingData); if (unlikely(!__pyx_tuple__61)) __PYX_ERR(0, 1096, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__61);
+  __Pyx_GIVEREF(__pyx_tuple__61);
+  __pyx_codeobj__62 = (PyObject*)__Pyx_PyCode_New(3, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__61, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classify, 1096, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__62)) __PYX_ERR(0, 1096, __pyx_L1_error)
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_classify, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__62)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1096, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__63);
+  __pyx_tuple__63 = PyTuple_Pack(1, ((PyObject *)Py_False)); if (unlikely(!__pyx_tuple__63)) __PYX_ERR(0, 1096, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__63);
+  __Pyx_GIVEREF(__pyx_tuple__63);
+/* … */
+struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify {
+  PyObject_HEAD
+  PyObject *__pyx_v_classifyWithMissingData;
+  PyObject *__pyx_v_classifyWithoutMissingData;
+};
+
+
 1097:         '''
+
 1098:         Classifies (prediction) the observations according to the tree.
+
 1099:
+
 1100:         Args
+
 1101:         ----
+
 1102:         observations : list of list
+
 1103:             The internal data format for the training data (combining X, Y, treatment).
+
 1104:
+
 1105:         dataMissing: boolean, optional (default = False)
+
 1106:             An indicator for if data are missing or not.
+
 1107:
+
 1108:         Returns
+
 1109:         -------
+
 1110:         tree.results, tree.upliftScore :
+
 1111:             The results in the leaf node.
+
 1112:         '''
+
 1113:
+
+1114:         def classifyWithoutMissingData(observations, tree):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData[] = "\n            Classifies (prediction) the observations according to the tree, assuming without missing data.\n\n            Args\n            ----\n            observations : list of list\n                The internal data format for the training data (combining X, Y, treatment).\n\n            Returns\n            -------\n            tree.results, tree.upliftScore :\n                The results in the leaf node.\n            ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData = {"classifyWithoutMissingData", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_observations = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classifyWithoutMissingData (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,0};
+    PyObject* values[2] = {0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("classifyWithoutMissingData", 1, 2, 2, 1); __PYX_ERR(0, 1114, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classifyWithoutMissingData") < 0)) __PYX_ERR(0, 1114, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+    }
+    __pyx_v_observations = values[0];
+    __pyx_v_tree = values[1];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("classifyWithoutMissingData", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1114, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithoutMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_self, __pyx_v_observations, __pyx_v_tree);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_outer_scope;
+  PyObject *__pyx_v_v = NULL;
+  PyObject *__pyx_v_branch = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classifyWithoutMissingData", 0);
+  __pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *) __Pyx_CyFunction_GetClosure(__pyx_self);
+  __pyx_cur_scope = __pyx_outer_scope;
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithoutMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_v);
+  __Pyx_XDECREF(__pyx_v_branch);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__8 = PyTuple_Pack(4, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_v, __pyx_n_s_branch); if (unlikely(!__pyx_tuple__8)) __PYX_ERR(0, 1114, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__8);
+  __Pyx_GIVEREF(__pyx_tuple__8);
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData, 0, __pyx_n_s_UpliftTreeClassifier_classify_lo, ((PyObject*)__pyx_cur_scope), __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__9)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1114, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GIVEREF(__pyx_t_1);
+  __pyx_cur_scope->__pyx_v_classifyWithoutMissingData = __pyx_t_1;
+  __pyx_t_1 = 0;
+  __pyx_codeobj__9 = (PyObject*)__Pyx_PyCode_New(2, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__8, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classifyWithoutMissingData, 1114, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__9)) __PYX_ERR(0, 1114, __pyx_L1_error)
+
 1115:             '''
+
 1116:             Classifies (prediction) the observations according to the tree, assuming without missing data.
+
 1117:
+
 1118:             Args
+
 1119:             ----
+
 1120:             observations : list of list
+
 1121:                 The internal data format for the training data (combining X, Y, treatment).
+
 1122:
+
 1123:             Returns
+
 1124:             -------
+
 1125:             tree.results, tree.upliftScore :
+
 1126:                 The results in the leaf node.
+
 1127:             '''
+
+1128:             if tree.results is not None:  # leaf
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1128, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = (__pyx_t_1 != Py_None);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_3 = (__pyx_t_2 != 0);
+  if (__pyx_t_3) {
+/* … */
+  }
+
+1129:                 return tree.results, tree.upliftScore
+
    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1129, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_upliftScore); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1129, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1129, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_GIVEREF(__pyx_t_1);
+    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_1);
+    __Pyx_GIVEREF(__pyx_t_4);
+    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_t_4);
+    __pyx_t_1 = 0;
+    __pyx_t_4 = 0;
+    __pyx_r = __pyx_t_5;
+    __pyx_t_5 = 0;
+    goto __pyx_L0;
+
 1130:             else:
+
+1131:                 v = observations[tree.col]
+
  /*else*/ {
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1131, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_observations, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1131, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_v_v = __pyx_t_4;
+    __pyx_t_4 = 0;
+
+1132:                 branch = None
+
    __Pyx_INCREF(Py_None);
+    __pyx_v_branch = Py_None;
+
+1133:                 if isinstance(v, int) or isinstance(v, float):
+
    __pyx_t_2 = PyInt_Check(__pyx_v_v); 
+    __pyx_t_6 = (__pyx_t_2 != 0);
+    if (!__pyx_t_6) {
+    } else {
+      __pyx_t_3 = __pyx_t_6;
+      goto __pyx_L5_bool_binop_done;
+    }
+    __pyx_t_6 = PyFloat_Check(__pyx_v_v); 
+    __pyx_t_2 = (__pyx_t_6 != 0);
+    __pyx_t_3 = __pyx_t_2;
+    __pyx_L5_bool_binop_done:;
+    if (__pyx_t_3) {
+/* … */
+      goto __pyx_L4;
+    }
+
+1134:                     if v >= tree.value:
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1134, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_5 = PyObject_RichCompare(__pyx_v_v, __pyx_t_4, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1134, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1134, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (__pyx_t_3) {
+/* … */
+        goto __pyx_L7;
+      }
+
+1135:                         branch = tree.trueBranch
+
        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1135, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_5);
+        __pyx_t_5 = 0;
+
 1136:                     else:
+
+1137:                         branch = tree.falseBranch
+
      /*else*/ {
+        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1137, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_5);
+        __pyx_t_5 = 0;
+      }
+      __pyx_L7:;
+
 1138:                 else:
+
+1139:                     if v == tree.value:
+
    /*else*/ {
+      __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1139, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __pyx_t_4 = PyObject_RichCompare(__pyx_v_v, __pyx_t_5, Py_EQ); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1139, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1139, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      if (__pyx_t_3) {
+/* … */
+        goto __pyx_L8;
+      }
+
+1140:                         branch = tree.trueBranch
+
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1140, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
+        __pyx_t_4 = 0;
+
 1141:                     else:
+
+1142:                         branch = tree.falseBranch
+
      /*else*/ {
+        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1142, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
+        __pyx_t_4 = 0;
+      }
+      __pyx_L8:;
+    }
+    __pyx_L4:;
+  }
+
+1143:             return classifyWithoutMissingData(observations, branch)
+
  __Pyx_XDECREF(__pyx_r);
+  if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithoutMissingData)) { __Pyx_RaiseClosureNameError("classifyWithoutMissingData"); __PYX_ERR(0, 1143, __pyx_L1_error) }
+  __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_cur_scope->__pyx_v_classifyWithoutMissingData, __pyx_v_observations, __pyx_v_branch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1143, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_r = __pyx_t_4;
+  __pyx_t_4 = 0;
+  goto __pyx_L0;
+
 1144:
+
+1145:         def classifyWithMissingData(observations, tree):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData[] = "\n            Classifies (prediction) the observations according to the tree, assuming with missing data.\n\n            Args\n            ----\n            observations : list of list\n                The internal data format for the training data (combining X, Y, treatment).\n\n            Returns\n            -------\n            tree.results, tree.upliftScore :\n                The results in the leaf node.\n            ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData = {"classifyWithMissingData", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_observations = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classifyWithMissingData (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,0};
+    PyObject* values[2] = {0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("classifyWithMissingData", 1, 2, 2, 1); __PYX_ERR(0, 1145, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classifyWithMissingData") < 0)) __PYX_ERR(0, 1145, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+    }
+    __pyx_v_observations = values[0];
+    __pyx_v_tree = values[1];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("classifyWithMissingData", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1145, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_self, __pyx_v_observations, __pyx_v_tree);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_outer_scope;
+  PyObject *__pyx_v_v = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_v_fr = NULL;
+  PyObject *__pyx_v_tcount = NULL;
+  PyObject *__pyx_v_fcount = NULL;
+  PyObject *__pyx_v_tw = NULL;
+  PyObject *__pyx_v_fw = NULL;
+  PyObject *__pyx_v_result = NULL;
+  PyObject *__pyx_v_k = NULL;
+  PyObject *__pyx_v_branch = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("classifyWithMissingData", 0);
+  __pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *) __Pyx_CyFunction_GetClosure(__pyx_self);
+  __pyx_cur_scope = __pyx_outer_scope;
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_v);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XDECREF(__pyx_v_fr);
+  __Pyx_XDECREF(__pyx_v_tcount);
+  __Pyx_XDECREF(__pyx_v_fcount);
+  __Pyx_XDECREF(__pyx_v_tw);
+  __Pyx_XDECREF(__pyx_v_fw);
+  __Pyx_XDECREF(__pyx_v_result);
+  __Pyx_XDECREF(__pyx_v_k);
+  __Pyx_XDECREF(__pyx_v_branch);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__10 = PyTuple_Pack(12, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_v, __pyx_n_s_tr, __pyx_n_s_fr, __pyx_n_s_tcount, __pyx_n_s_fcount, __pyx_n_s_tw, __pyx_n_s_fw, __pyx_n_s_result, __pyx_n_s_k, __pyx_n_s_branch); if (unlikely(!__pyx_tuple__10)) __PYX_ERR(0, 1145, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__10);
+  __Pyx_GIVEREF(__pyx_tuple__10);
+/* … */
+  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData, 0, __pyx_n_s_UpliftTreeClassifier_classify_lo_2, ((PyObject*)__pyx_cur_scope), __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__11)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1145, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GIVEREF(__pyx_t_1);
+  __pyx_cur_scope->__pyx_v_classifyWithMissingData = __pyx_t_1;
+  __pyx_t_1 = 0;
+  __pyx_codeobj__11 = (PyObject*)__Pyx_PyCode_New(2, 0, 12, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__10, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classifyWithMissingData, 1145, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__11)) __PYX_ERR(0, 1145, __pyx_L1_error)
+
 1146:             '''
+
 1147:             Classifies (prediction) the observations according to the tree, assuming with missing data.
+
 1148:
+
 1149:             Args
+
 1150:             ----
+
 1151:             observations : list of list
+
 1152:                 The internal data format for the training data (combining X, Y, treatment).
+
 1153:
+
 1154:             Returns
+
 1155:             -------
+
 1156:             tree.results, tree.upliftScore :
+
 1157:                 The results in the leaf node.
+
 1158:             '''
+
+1159:             if tree.results is not None:  # leaf
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1159, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = (__pyx_t_1 != Py_None);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_3 = (__pyx_t_2 != 0);
+  if (__pyx_t_3) {
+/* … */
+  }
+
+1160:                 return tree.results
+
    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1160, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_r = __pyx_t_1;
+    __pyx_t_1 = 0;
+    goto __pyx_L0;
+
 1161:             else:
+
+1162:                 v = observations[tree.col]
+
  /*else*/ {
+    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1162, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_observations, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1162, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_v_v = __pyx_t_4;
+    __pyx_t_4 = 0;
+
+1163:                 if v is None:
+
    __pyx_t_3 = (__pyx_v_v == Py_None);
+    __pyx_t_2 = (__pyx_t_3 != 0);
+    if (__pyx_t_2) {
+/* … */
+    }
+
+1164:                     tr = classifyWithMissingData(observations, tree.trueBranch)
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1164, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1164, __pyx_L1_error) }
+      __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1164, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_v_tr = __pyx_t_1;
+      __pyx_t_1 = 0;
+
+1165:                     fr = classifyWithMissingData(observations, tree.falseBranch)
+
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1165, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1165, __pyx_L1_error) }
+      __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1165, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_v_fr = __pyx_t_4;
+      __pyx_t_4 = 0;
+
+1166:                     tcount = sum(tr.values())
+
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tr, __pyx_n_s_values); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1166, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_5 = NULL;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
+        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
+        if (likely(__pyx_t_5)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+          __Pyx_INCREF(__pyx_t_5);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_1, function);
+        }
+      }
+      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
+      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1166, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1166, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_v_tcount = __pyx_t_1;
+      __pyx_t_1 = 0;
+
+1167:                     fcount = sum(fr.values())
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_fr, __pyx_n_s_values); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1167, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_5 = NULL;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
+        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_4);
+        if (likely(__pyx_t_5)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+          __Pyx_INCREF(__pyx_t_5);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_4, function);
+        }
+      }
+      __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
+      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1167, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1167, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_v_fcount = __pyx_t_4;
+      __pyx_t_4 = 0;
+
+1168:                     tw = float(tcount) / (tcount + fcount)
+
      __pyx_t_4 = __Pyx_PyNumber_Float(__pyx_v_tcount); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1168, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_1 = PyNumber_Add(__pyx_v_tcount, __pyx_v_fcount); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1168, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1168, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_v_tw = __pyx_t_5;
+      __pyx_t_5 = 0;
+
+1169:                     fw = float(fcount) / (tcount + fcount)
+
      __pyx_t_5 = __Pyx_PyNumber_Float(__pyx_v_fcount); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1169, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __pyx_t_1 = PyNumber_Add(__pyx_v_tcount, __pyx_v_fcount); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1169, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_4 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1169, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_v_fw = __pyx_t_4;
+      __pyx_t_4 = 0;
+
 1170:
+
 1171:                     # Problem description: http://blog.ludovf.net/python-collections-defaultdict/
+
+1172:                     result = defaultdict(int)
+
      __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1172, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_5 = NULL;
+      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
+        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
+        if (likely(__pyx_t_5)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+          __Pyx_INCREF(__pyx_t_5);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_1, function);
+        }
+      }
+      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_5, ((PyObject *)(&PyInt_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_1, ((PyObject *)(&PyInt_Type)));
+      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1172, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      __pyx_v_result = __pyx_t_4;
+      __pyx_t_4 = 0;
+
+1173:                     for k, v in tr.items():
+
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tr, __pyx_n_s_items); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __pyx_t_5 = NULL;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
+        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
+        if (likely(__pyx_t_5)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
+          __Pyx_INCREF(__pyx_t_5);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_1, function);
+        }
+      }
+      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
+      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
+        __pyx_t_1 = __pyx_t_4; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
+        __pyx_t_7 = NULL;
+      } else {
+        __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1173, __pyx_L1_error)
+      }
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      for (;;) {
+        if (likely(!__pyx_t_7)) {
+          if (likely(PyList_CheckExact(__pyx_t_1))) {
+            if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_4 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1173, __pyx_L1_error)
+            #else
+            __pyx_t_4 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_4);
+            #endif
+          } else {
+            if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1173, __pyx_L1_error)
+            #else
+            __pyx_t_4 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_4);
+            #endif
+          }
+        } else {
+          __pyx_t_4 = __pyx_t_7(__pyx_t_1);
+          if (unlikely(!__pyx_t_4)) {
+            PyObject* exc_type = PyErr_Occurred();
+            if (exc_type) {
+              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+              else __PYX_ERR(0, 1173, __pyx_L1_error)
+            }
+            break;
+          }
+          __Pyx_GOTREF(__pyx_t_4);
+        }
+        if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
+          PyObject* sequence = __pyx_t_4;
+          Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+          if (unlikely(size != 2)) {
+            if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+            else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+            __PYX_ERR(0, 1173, __pyx_L1_error)
+          }
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          if (likely(PyTuple_CheckExact(sequence))) {
+            __pyx_t_5 = PyTuple_GET_ITEM(sequence, 0); 
+            __pyx_t_8 = PyTuple_GET_ITEM(sequence, 1); 
+          } else {
+            __pyx_t_5 = PyList_GET_ITEM(sequence, 0); 
+            __pyx_t_8 = PyList_GET_ITEM(sequence, 1); 
+          }
+          __Pyx_INCREF(__pyx_t_5);
+          __Pyx_INCREF(__pyx_t_8);
+          #else
+          __pyx_t_5 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1173, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_5);
+          __pyx_t_8 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1173, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_8);
+          #endif
+          __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        } else {
+          Py_ssize_t index = -1;
+          __pyx_t_9 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1173, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+          __pyx_t_10 = Py_TYPE(__pyx_t_9)->tp_iternext;
+          index = 0; __pyx_t_5 = __pyx_t_10(__pyx_t_9); if (unlikely(!__pyx_t_5)) goto __pyx_L7_unpacking_failed;
+          __Pyx_GOTREF(__pyx_t_5);
+          index = 1; __pyx_t_8 = __pyx_t_10(__pyx_t_9); if (unlikely(!__pyx_t_8)) goto __pyx_L7_unpacking_failed;
+          __Pyx_GOTREF(__pyx_t_8);
+          if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_9), 2) < 0) __PYX_ERR(0, 1173, __pyx_L1_error)
+          __pyx_t_10 = NULL;
+          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+          goto __pyx_L8_unpacking_done;
+          __pyx_L7_unpacking_failed:;
+          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+          __pyx_t_10 = NULL;
+          if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+          __PYX_ERR(0, 1173, __pyx_L1_error)
+          __pyx_L8_unpacking_done:;
+        }
+        __Pyx_XDECREF_SET(__pyx_v_k, __pyx_t_5);
+        __pyx_t_5 = 0;
+        __Pyx_DECREF_SET(__pyx_v_v, __pyx_t_8);
+        __pyx_t_8 = 0;
+/* … */
+      }
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+1174:                         result[k] += v * tw
+
        __Pyx_INCREF(__pyx_v_k);
+        __pyx_t_4 = __pyx_v_k;
+        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_result, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1174, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __pyx_t_5 = PyNumber_Multiply(__pyx_v_v, __pyx_v_tw); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1174, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_8, __pyx_t_5); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1174, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        if (unlikely(PyObject_SetItem(__pyx_v_result, __pyx_t_4, __pyx_t_9) < 0)) __PYX_ERR(0, 1174, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+1175:                     for k, v in fr.items():
+
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_fr, __pyx_n_s_items); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1175, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_9 = NULL;
+      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
+        __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_4);
+        if (likely(__pyx_t_9)) {
+          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
+          __Pyx_INCREF(__pyx_t_9);
+          __Pyx_INCREF(function);
+          __Pyx_DECREF_SET(__pyx_t_4, function);
+        }
+      }
+      __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_9) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
+      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
+      if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
+        __pyx_t_4 = __pyx_t_1; __Pyx_INCREF(__pyx_t_4); __pyx_t_6 = 0;
+        __pyx_t_7 = NULL;
+      } else {
+        __pyx_t_6 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1175, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __pyx_t_7 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1175, __pyx_L1_error)
+      }
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      for (;;) {
+        if (likely(!__pyx_t_7)) {
+          if (likely(PyList_CheckExact(__pyx_t_4))) {
+            if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_4)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_1 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_6); __Pyx_INCREF(__pyx_t_1); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1175, __pyx_L1_error)
+            #else
+            __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_1);
+            #endif
+          } else {
+            if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
+            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+            __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_6); __Pyx_INCREF(__pyx_t_1); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1175, __pyx_L1_error)
+            #else
+            __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
+            __Pyx_GOTREF(__pyx_t_1);
+            #endif
+          }
+        } else {
+          __pyx_t_1 = __pyx_t_7(__pyx_t_4);
+          if (unlikely(!__pyx_t_1)) {
+            PyObject* exc_type = PyErr_Occurred();
+            if (exc_type) {
+              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+              else __PYX_ERR(0, 1175, __pyx_L1_error)
+            }
+            break;
+          }
+          __Pyx_GOTREF(__pyx_t_1);
+        }
+        if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
+          PyObject* sequence = __pyx_t_1;
+          Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
+          if (unlikely(size != 2)) {
+            if (size > 2) __Pyx_RaiseTooManyValuesError(2);
+            else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
+            __PYX_ERR(0, 1175, __pyx_L1_error)
+          }
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          if (likely(PyTuple_CheckExact(sequence))) {
+            __pyx_t_9 = PyTuple_GET_ITEM(sequence, 0); 
+            __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
+          } else {
+            __pyx_t_9 = PyList_GET_ITEM(sequence, 0); 
+            __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
+          }
+          __Pyx_INCREF(__pyx_t_9);
+          __Pyx_INCREF(__pyx_t_5);
+          #else
+          __pyx_t_9 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1175, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_9);
+          __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1175, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_5);
+          #endif
+          __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+        } else {
+          Py_ssize_t index = -1;
+          __pyx_t_8 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1175, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_8);
+          __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+          __pyx_t_10 = Py_TYPE(__pyx_t_8)->tp_iternext;
+          index = 0; __pyx_t_9 = __pyx_t_10(__pyx_t_8); if (unlikely(!__pyx_t_9)) goto __pyx_L11_unpacking_failed;
+          __Pyx_GOTREF(__pyx_t_9);
+          index = 1; __pyx_t_5 = __pyx_t_10(__pyx_t_8); if (unlikely(!__pyx_t_5)) goto __pyx_L11_unpacking_failed;
+          __Pyx_GOTREF(__pyx_t_5);
+          if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_8), 2) < 0) __PYX_ERR(0, 1175, __pyx_L1_error)
+          __pyx_t_10 = NULL;
+          __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+          goto __pyx_L12_unpacking_done;
+          __pyx_L11_unpacking_failed:;
+          __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+          __pyx_t_10 = NULL;
+          if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
+          __PYX_ERR(0, 1175, __pyx_L1_error)
+          __pyx_L12_unpacking_done:;
+        }
+        __Pyx_XDECREF_SET(__pyx_v_k, __pyx_t_9);
+        __pyx_t_9 = 0;
+        __Pyx_DECREF_SET(__pyx_v_v, __pyx_t_5);
+        __pyx_t_5 = 0;
+/* … */
+      }
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+1176:                         result[k] += v * fw
+
        __Pyx_INCREF(__pyx_v_k);
+        __pyx_t_1 = __pyx_v_k;
+        __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_result, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1176, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        __pyx_t_9 = PyNumber_Multiply(__pyx_v_v, __pyx_v_fw); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1176, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_9);
+        __pyx_t_8 = PyNumber_InPlaceAdd(__pyx_t_5, __pyx_t_9); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1176, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_8);
+        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+        if (unlikely(PyObject_SetItem(__pyx_v_result, __pyx_t_1, __pyx_t_8) < 0)) __PYX_ERR(0, 1176, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+1177:                     return dict(result)
+
      __Pyx_XDECREF(__pyx_r);
+      __pyx_t_4 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyDict_Type)), __pyx_v_result); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1177, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_r = __pyx_t_4;
+      __pyx_t_4 = 0;
+      goto __pyx_L0;
+
 1178:                 else:
+
+1179:                     branch = None
+
    /*else*/ {
+      __Pyx_INCREF(Py_None);
+      __pyx_v_branch = Py_None;
+
+1180:                     if isinstance(v, int) or isinstance(v, float):
+
      __pyx_t_3 = PyInt_Check(__pyx_v_v); 
+      __pyx_t_11 = (__pyx_t_3 != 0);
+      if (!__pyx_t_11) {
+      } else {
+        __pyx_t_2 = __pyx_t_11;
+        goto __pyx_L14_bool_binop_done;
+      }
+      __pyx_t_11 = PyFloat_Check(__pyx_v_v); 
+      __pyx_t_3 = (__pyx_t_11 != 0);
+      __pyx_t_2 = __pyx_t_3;
+      __pyx_L14_bool_binop_done:;
+      if (__pyx_t_2) {
+/* … */
+        goto __pyx_L13;
+      }
+
+1181:                         if v >= tree.value:
+
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1181, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        __pyx_t_1 = PyObject_RichCompare(__pyx_v_v, __pyx_t_4, Py_GE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1181, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1181, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+        if (__pyx_t_2) {
+/* … */
+          goto __pyx_L16;
+        }
+
+1182:                             branch = tree.trueBranch
+
          __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1182, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_1);
+          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_1);
+          __pyx_t_1 = 0;
+
 1183:                         else:
+
+1184:                             branch = tree.falseBranch
+
        /*else*/ {
+          __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1184, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_1);
+          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_1);
+          __pyx_t_1 = 0;
+        }
+        __pyx_L16:;
+
 1185:                     else:
+
+1186:                         if v == tree.value:
+
      /*else*/ {
+        __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1186, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        __pyx_t_4 = PyObject_RichCompare(__pyx_v_v, __pyx_t_1, Py_EQ); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1186, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1186, __pyx_L1_error)
+        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+        if (__pyx_t_2) {
+/* … */
+          goto __pyx_L17;
+        }
+
+1187:                             branch = tree.trueBranch
+
          __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1187, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
+          __pyx_t_4 = 0;
+
 1188:                         else:
+
+1189:                             branch = tree.falseBranch
+
        /*else*/ {
+          __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1189, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_4);
+          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
+          __pyx_t_4 = 0;
+        }
+        __pyx_L17:;
+      }
+      __pyx_L13:;
+    }
+
+1190:                 return classifyWithMissingData(observations, branch)
+
    __Pyx_XDECREF(__pyx_r);
+    if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1190, __pyx_L1_error) }
+    __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_v_branch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1190, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_r = __pyx_t_4;
+    __pyx_t_4 = 0;
+    goto __pyx_L0;
+  }
+
 1191:
+
 1192:         # function body
+
+1193:         if dataMissing:
+
  __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_v_dataMissing); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1193, __pyx_L1_error)
+  if (__pyx_t_2) {
+/* … */
+  }
+
+1194:             return classifyWithMissingData(observations, tree)
+
    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_v_tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1194, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_r = __pyx_t_1;
+    __pyx_t_1 = 0;
+    goto __pyx_L0;
+
 1195:         else:
+
+1196:             return classifyWithoutMissingData(observations, tree)
+
  /*else*/ {
+    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_cur_scope->__pyx_v_classifyWithoutMissingData, __pyx_v_observations, __pyx_v_tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1196, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_r = __pyx_t_1;
+    __pyx_t_1 = 0;
+    goto __pyx_L0;
+  }
+
 1197:
+
 1198:
+
 1199: # Uplift Random Forests
+
+1200: class UpliftRandomForestClassifier:
+
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_UpliftRandomForestClassifier, __pyx_n_s_UpliftRandomForestClassifier, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Uplift_Random_Forest_for_Classi); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1200, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_UpliftRandomForestClassifier, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1200, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_d, __pyx_n_s_UpliftRandomForestClassifier, __pyx_t_2) < 0) __PYX_ERR(0, 1200, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 1201:     """ Uplift Random Forest for Classification Task.
+
 1202:
+
 1203:     Parameters
+
 1204:     ----------
+
 1205:     n_estimators : integer, optional (default=10)
+
 1206:         The number of trees in the uplift random forest.
+
 1207:
+
 1208:     evaluationFunction : string
+
 1209:         Choose from one of the models: 'KL', 'ED', 'Chi', 'CTS', 'DDP'.
+
 1210:
+
 1211:     max_features: int, optional (default=10)
+
 1212:         The number of features to consider when looking for the best split.
+
 1213:
+
 1214:     random_state: int, RandomState instance or None (default=None)
+
 1215:         A random seed or `np.random.RandomState` to control randomness in building the trees and forest.
+
 1216:
+
 1217:     max_depth: int, optional (default=5)
+
 1218:         The maximum depth of the tree.
+
 1219:
+
 1220:     min_samples_leaf: int, optional (default=100)
+
 1221:         The minimum number of samples required to be split at a leaf node.
+
 1222:
+
 1223:     min_samples_treatment: int, optional (default=10)
+
 1224:         The minimum number of samples required of the experiment group to be split at a leaf node.
+
 1225:
+
 1226:     n_reg: int, optional (default=10)
+
 1227:         The regularization parameter defined in Rzepakowski et al. 2012, the
+
 1228:         weight (in terms of sample size) of the parent node influence on the
+
 1229:         child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
+
 1230:
+
 1231:     control_name: string
+
 1232:         The name of the control group (other experiment groups will be regarded as treatment groups)
+
 1233:
+
 1234:     normalization: boolean, optional (default=True)
+
 1235:         The normalization factor defined in Rzepakowski et al. 2012,
+
 1236:         correcting for tests with large number of splits and imbalanced
+
 1237:         treatment and control splits
+
 1238:
+
 1239:     n_jobs: int, optional (default=-1)
+
 1240:         The parallelization parameter to define how many parallel jobs need to be created.
+
 1241:         This is passed on to joblib library for parallelizing uplift-tree creation and prediction.
+
 1242:
+
 1243:     joblib_prefer: str, optional (default="threads")
+
 1244:         The preferred backend for joblib (passed as `prefer` to joblib.Parallel). See the joblib
+
 1245:         documentation for valid values.
+
 1246:
+
 1247:     Outputs
+
 1248:     ----------
+
 1249:     df_res: pandas dataframe
+
 1250:         A user-level results dataframe containing the estimated individual treatment effect.
+
 1251:     """
+
+1252:     def __init__(self,
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__[] = "\n        Initialize the UpliftRandomForestClassifier class.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_control_name = 0;
+  PyObject *__pyx_v_n_estimators = 0;
+  PyObject *__pyx_v_max_features = 0;
+  PyObject *__pyx_v_random_state = 0;
+  PyObject *__pyx_v_max_depth = 0;
+  PyObject *__pyx_v_min_samples_leaf = 0;
+  PyObject *__pyx_v_min_samples_treatment = 0;
+  PyObject *__pyx_v_n_reg = 0;
+  PyObject *__pyx_v_evaluationFunction = 0;
+  PyObject *__pyx_v_normalization = 0;
+  PyObject *__pyx_v_n_jobs = 0;
+  PyObject *__pyx_v_joblib_prefer = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_control_name,&__pyx_n_s_n_estimators,&__pyx_n_s_max_features,&__pyx_n_s_random_state,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_evaluationFunction,&__pyx_n_s_normalization,&__pyx_n_s_n_jobs,&__pyx_n_s_joblib_prefer,0};
+    PyObject* values[13] = {0,0,0,0,0,0,0,0,0,0,0,0,0};
+    values[2] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[3] = ((PyObject *)((PyObject *)__pyx_int_10));
+/* … */
+  /* function exit code */
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_control_name, PyObject *__pyx_v_n_estimators, PyObject *__pyx_v_max_features, PyObject *__pyx_v_random_state, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_evaluationFunction, PyObject *__pyx_v_normalization, PyObject *__pyx_v_n_jobs, PyObject *__pyx_v_joblib_prefer) {
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("__init__", 0);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__64 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_control_name, __pyx_n_s_n_estimators, __pyx_n_s_max_features, __pyx_n_s_random_state, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_evaluationFunction, __pyx_n_s_normalization, __pyx_n_s_n_jobs, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_tuple__64)) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__64);
+  __Pyx_GIVEREF(__pyx_tuple__64);
+  __pyx_codeobj__65 = (PyObject*)__Pyx_PyCode_New(13, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__64, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 1252, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__65)) __PYX_ERR(0, 1252, __pyx_L1_error)
+/* … */
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_joblib_prefer, __pyx_n_u_str) < 0) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__, 0, __pyx_n_s_UpliftRandomForestClassifier___i, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__65)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__66);
+  __Pyx_CyFunction_SetAnnotationsDict(__pyx_t_5, __pyx_t_2);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_5) < 0) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_tuple__66 = PyTuple_Pack(11, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_10), ((PyObject *)Py_None), ((PyObject *)__pyx_int_5), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_10), ((PyObject*)__pyx_n_s_KL), ((PyObject *)Py_True), ((PyObject *)__pyx_int_neg_1), ((PyObject*)__pyx_n_s_threads)); if (unlikely(!__pyx_tuple__66)) __PYX_ERR(0, 1252, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__66);
+  __Pyx_GIVEREF(__pyx_tuple__66);
+
 1253:                  control_name,
+
 1254:                  n_estimators=10,
+
 1255:                  max_features=10,
+
+1256:                  random_state=None,
+
    values[4] = ((PyObject *)((PyObject *)Py_None));
+    values[5] = ((PyObject *)((PyObject *)__pyx_int_5));
+    values[6] = ((PyObject *)((PyObject *)__pyx_int_100));
+    values[7] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[8] = ((PyObject *)((PyObject *)__pyx_int_10));
+    values[9] = ((PyObject *)((PyObject*)__pyx_n_s_KL));
+
 1257:                  max_depth=5,
+
 1258:                  min_samples_leaf=100,
+
 1259:                  min_samples_treatment=10,
+
 1260:                  n_reg=10,
+
 1261:                  evaluationFunction='KL',
+
+1262:                  normalization=True,
+
    values[10] = ((PyObject *)((PyObject *)Py_True));
+    values[11] = ((PyObject *)((PyObject *)__pyx_int_neg_1));
+    values[12] = ((PyObject*)((PyObject*)__pyx_n_s_threads));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
+        CYTHON_FALLTHROUGH;
+        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
+        CYTHON_FALLTHROUGH;
+        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
+        CYTHON_FALLTHROUGH;
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 13, 1); __PYX_ERR(0, 1252, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_estimators);
+          if (value) { values[2] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_features);
+          if (value) { values[3] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  4:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
+          if (value) { values[4] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  5:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
+          if (value) { values[5] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  6:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
+          if (value) { values[6] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  7:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
+          if (value) { values[7] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  8:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
+          if (value) { values[8] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case  9:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_evaluationFunction);
+          if (value) { values[9] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 10:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_normalization);
+          if (value) { values[10] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 11:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_jobs);
+          if (value) { values[11] = value; kw_args--; }
+        }
+        CYTHON_FALLTHROUGH;
+        case 12:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_joblib_prefer);
+          if (value) { values[12] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1252, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
+        CYTHON_FALLTHROUGH;
+        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
+        CYTHON_FALLTHROUGH;
+        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
+        CYTHON_FALLTHROUGH;
+        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
+        CYTHON_FALLTHROUGH;
+        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
+        CYTHON_FALLTHROUGH;
+        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
+        CYTHON_FALLTHROUGH;
+        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
+        CYTHON_FALLTHROUGH;
+        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
+        CYTHON_FALLTHROUGH;
+        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
+        CYTHON_FALLTHROUGH;
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_control_name = values[1];
+    __pyx_v_n_estimators = values[2];
+    __pyx_v_max_features = values[3];
+    __pyx_v_random_state = values[4];
+    __pyx_v_max_depth = values[5];
+    __pyx_v_min_samples_leaf = values[6];
+    __pyx_v_min_samples_treatment = values[7];
+    __pyx_v_n_reg = values[8];
+    __pyx_v_evaluationFunction = values[9];
+    __pyx_v_normalization = values[10];
+    __pyx_v_n_jobs = values[11];
+    __pyx_v_joblib_prefer = ((PyObject*)values[12]);
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 13, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1252, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_joblib_prefer), (&PyString_Type), 1, "joblib_prefer", 1))) __PYX_ERR(0, 1264, __pyx_L1_error)
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__(__pyx_self, __pyx_v_self, __pyx_v_control_name, __pyx_v_n_estimators, __pyx_v_max_features, __pyx_v_random_state, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_evaluationFunction, __pyx_v_normalization, __pyx_v_n_jobs, __pyx_v_joblib_prefer);
+
 1263:                  n_jobs=-1,
+
 1264:                  joblib_prefer: str = "threads"):
+
 1265:
+
 1266:         """
+
 1267:         Initialize the UpliftRandomForestClassifier class.
+
 1268:         """
+
+1269:         self.n_estimators = n_estimators
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_estimators, __pyx_v_n_estimators) < 0) __PYX_ERR(0, 1269, __pyx_L1_error)
+
+1270:         self.max_features = max_features
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_features, __pyx_v_max_features) < 0) __PYX_ERR(0, 1270, __pyx_L1_error)
+
+1271:         self.random_state = random_state
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 1271, __pyx_L1_error)
+
+1272:         self.max_depth = max_depth
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 1272, __pyx_L1_error)
+
+1273:         self.min_samples_leaf = min_samples_leaf
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 1273, __pyx_L1_error)
+
+1274:         self.min_samples_treatment = min_samples_treatment
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1274, __pyx_L1_error)
+
+1275:         self.n_reg = n_reg
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1275, __pyx_L1_error)
+
+1276:         self.evaluationFunction = evaluationFunction
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_v_evaluationFunction) < 0) __PYX_ERR(0, 1276, __pyx_L1_error)
+
+1277:         self.control_name = control_name
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 1277, __pyx_L1_error)
+
+1278:         self.normalization = normalization
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_normalization, __pyx_v_normalization) < 0) __PYX_ERR(0, 1278, __pyx_L1_error)
+
+1279:         self.n_jobs = n_jobs
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs, __pyx_v_n_jobs) < 0) __PYX_ERR(0, 1279, __pyx_L1_error)
+
+1280:         self.joblib_prefer = joblib_prefer
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_joblib_prefer, __pyx_v_joblib_prefer) < 0) __PYX_ERR(0, 1280, __pyx_L1_error)
+
 1281:
+
+1282:         assert control_name is not None and isinstance(control_name, str), \
+
  #ifndef CYTHON_WITHOUT_ASSERTIONS
+  if (unlikely(!Py_OptimizeFlag)) {
+    __pyx_t_2 = (__pyx_v_control_name != Py_None);
+    __pyx_t_3 = (__pyx_t_2 != 0);
+    if (__pyx_t_3) {
+    } else {
+      __pyx_t_1 = __pyx_t_3;
+      goto __pyx_L3_bool_binop_done;
+    }
+    __pyx_t_3 = PyString_Check(__pyx_v_control_name); 
+    __pyx_t_2 = (__pyx_t_3 != 0);
+    __pyx_t_1 = __pyx_t_2;
+    __pyx_L3_bool_binop_done:;
+    if (unlikely(!__pyx_t_1)) {
+
+1283:             f"control_group should be string but {control_name} is passed"
+
      __pyx_t_4 = PyTuple_New(3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1283, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_4);
+      __pyx_t_5 = 0;
+      __pyx_t_6 = 127;
+      __Pyx_INCREF(__pyx_kp_u_control_group_should_be_string_b);
+      __pyx_t_5 += 35;
+      __Pyx_GIVEREF(__pyx_kp_u_control_group_should_be_string_b);
+      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_kp_u_control_group_should_be_string_b);
+      __pyx_t_7 = __Pyx_PyObject_FormatSimple(__pyx_v_control_name, __pyx_empty_unicode); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1283, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __pyx_t_6 = (__Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) > __pyx_t_6) ? __Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) : __pyx_t_6;
+      __pyx_t_5 += __Pyx_PyUnicode_GET_LENGTH(__pyx_t_7);
+      __Pyx_GIVEREF(__pyx_t_7);
+      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_t_7);
+      __pyx_t_7 = 0;
+      __Pyx_INCREF(__pyx_kp_u_is_passed);
+      __pyx_t_5 += 10;
+      __Pyx_GIVEREF(__pyx_kp_u_is_passed);
+      PyTuple_SET_ITEM(__pyx_t_4, 2, __pyx_kp_u_is_passed);
+      __pyx_t_7 = __Pyx_PyUnicode_Join(__pyx_t_4, 3, __pyx_t_5, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1283, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+      PyErr_SetObject(PyExc_AssertionError, __pyx_t_7);
+      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+      __PYX_ERR(0, 1282, __pyx_L1_error)
+    }
+  }
+  #endif
+
+1284:         self.control_name = control_name
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 1284, __pyx_L1_error)
+
+1285:         self.classes_ = [control_name]
+
  __pyx_t_7 = PyList_New(1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1285, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_INCREF(__pyx_v_control_name);
+  __Pyx_GIVEREF(__pyx_v_control_name);
+  PyList_SET_ITEM(__pyx_t_7, 0, __pyx_v_control_name);
+  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_7) < 0) __PYX_ERR(0, 1285, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1286:         self.n_class = 1
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_int_1) < 0) __PYX_ERR(0, 1286, __pyx_L1_error)
+
 1287:
+
+1288:         if self.n_jobs == -1:
+
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_4 = __Pyx_PyInt_EqObjC(__pyx_t_7, __pyx_int_neg_1, -1L, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1288, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 1288, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  if (__pyx_t_1) {
+/* … */
+  }
+
+1289:             self.n_jobs = mp.cpu_count()
+
    __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_mp); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_cpu_count); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __pyx_t_7 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_8))) {
+      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_8);
+      if (likely(__pyx_t_7)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_8);
+        __Pyx_INCREF(__pyx_t_7);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_8, function);
+      }
+    }
+    __pyx_t_4 = (__pyx_t_7) ? __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_7) : __Pyx_PyObject_CallNoArg(__pyx_t_8);
+    __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+    if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1289, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs, __pyx_t_4) < 0) __PYX_ERR(0, 1289, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
 1290:
+
+1291:     def fit(self, X, treatment, y):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit[] = "\n        Fit the UpliftRandomForestClassifier.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 1291, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 1291, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 1291, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 1291, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_treatment = values[2];
+    __pyx_v_y = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1291, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value); /* proto */
+/* … */
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *__pyx_cur_scope;
+  PyObject *__pyx_v_random_state = NULL;
+  PyObject *__pyx_v_treatment_groups = NULL;
+  PyObject *__pyx_v_tr = NULL;
+  PyObject *__pyx_v_all_importances = NULL;
+  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
+  PyObject *__pyx_v_x = NULL;
+  PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator = 0;
+  PyObject *__pyx_v_tree = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("fit", 0);
+  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit, __pyx_empty_tuple, NULL);
+  if (unlikely(!__pyx_cur_scope)) {
+    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *)Py_None);
+    __Pyx_INCREF(Py_None);
+    __PYX_ERR(0, 1291, __pyx_L1_error)
+  } else {
+    __Pyx_GOTREF(__pyx_cur_scope);
+  }
+  __pyx_cur_scope->__pyx_v_self = __pyx_v_self;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_self);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_self);
+  __pyx_cur_scope->__pyx_v_X = __pyx_v_X;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_X);
+  __pyx_cur_scope->__pyx_v_treatment = __pyx_v_treatment;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_treatment);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_treatment);
+  __pyx_cur_scope->__pyx_v_y = __pyx_v_y;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_y);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_y);
+/* … */
+  /* function exit code */
+  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_XDECREF(__pyx_t_9);
+  __Pyx_XDECREF(__pyx_t_10);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_random_state);
+  __Pyx_XDECREF(__pyx_v_treatment_groups);
+  __Pyx_XDECREF(__pyx_v_tr);
+  __Pyx_XDECREF(__pyx_v_all_importances);
+  __Pyx_XDECREF(__pyx_v__);
+  __Pyx_XDECREF(__pyx_v_x);
+  __Pyx_XDECREF(__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator);
+  __Pyx_XDECREF(__pyx_v_tree);
+  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__67 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_random_state, __pyx_n_s_treatment_groups, __pyx_n_s_tr, __pyx_n_s_all_importances, __pyx_n_s__34, __pyx_n_s_x, __pyx_n_s_genexpr, __pyx_n_s_genexpr, __pyx_n_s_tree); if (unlikely(!__pyx_tuple__67)) __PYX_ERR(0, 1291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__67);
+  __Pyx_GIVEREF(__pyx_tuple__67);
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit, 0, __pyx_n_s_UpliftRandomForestClassifier_fit_2, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__68)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1291, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fit, __pyx_t_5) < 0) __PYX_ERR(0, 1291, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_codeobj__68 = (PyObject*)__Pyx_PyCode_New(4, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__67, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fit, 1291, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__68)) __PYX_ERR(0, 1291, __pyx_L1_error)
+/* … */
+struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit {
+  PyObject_HEAD
+  PyObject *__pyx_v_X;
+  PyObject *__pyx_v_self;
+  PyObject *__pyx_v_treatment;
+  PyObject *__pyx_v_y;
+};
+
+
 1292:         """
+
 1293:         Fit the UpliftRandomForestClassifier.
+
 1294:
+
 1295:         Args
+
 1296:         ----
+
 1297:         X : ndarray, shape = [num_samples, num_features]
+
 1298:             An ndarray of the covariates used to train the uplift model.
+
 1299:
+
 1300:         treatment : array-like, shape = [num_samples]
+
 1301:             An array containing the treatment group for each unit.
+
 1302:
+
 1303:         y : array-like, shape = [num_samples]
+
 1304:             An array containing the outcome of interest for each unit.
+
 1305:         """
+
+1306:         random_state = check_random_state(self.random_state)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1306, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_random_state = __pyx_t_1;
+  __pyx_t_1 = 0;
+
 1307:
+
 1308:         # Create forest
+
+1309:         self.uplift_forest = [
+
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1309, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+/* … */
+  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest, __pyx_t_1) < 0) __PYX_ERR(0, 1309, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+1310:             UpliftTreeClassifier(
+
    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_UpliftTreeClassifier); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1310, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+/* … */
+    __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1310, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 1309, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1311:                 max_features=self.max_features, max_depth=self.max_depth,
+
    __pyx_t_4 = __Pyx_PyDict_NewPresized(9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_max_features, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_max_depth, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1312:                 min_samples_leaf=self.min_samples_leaf,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1312, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_min_samples_leaf, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1313:                 min_samples_treatment=self.min_samples_treatment,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1313, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_min_samples_treatment, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1314:                 n_reg=self.n_reg,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1314, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_n_reg, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1315:                 evaluationFunction=self.evaluationFunction,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1315, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_evaluationFunction, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1316:                 control_name=self.control_name,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1316, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_control_name, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1317:                 normalization=self.normalization,
+
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_normalization); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1317, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_normalization, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1318:                 random_state=random_state.randint(MAX_INT))
+
    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_random_state, __pyx_n_s_randint); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1318, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_MAX_INT); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1318, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_9);
+    __pyx_t_10 = NULL;
+    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_8))) {
+      __pyx_t_10 = PyMethod_GET_SELF(__pyx_t_8);
+      if (likely(__pyx_t_10)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_8);
+        __Pyx_INCREF(__pyx_t_10);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_8, function);
+      }
+    }
+    __pyx_t_7 = (__pyx_t_10) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_10, __pyx_t_9) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_9);
+    __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
+    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
+    if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1318, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_random_state, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1319:             for _ in range(self.n_estimators)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_estimators); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1319, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
+    __pyx_t_2 = __pyx_t_3; __Pyx_INCREF(__pyx_t_2); __pyx_t_5 = 0;
+    __pyx_t_6 = NULL;
+  } else {
+    __pyx_t_5 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1319, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_6 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1319, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_6)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1319, __pyx_L1_error)
+        #else
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        #endif
+      } else {
+        if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1319, __pyx_L1_error)
+        #else
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        #endif
+      }
+    } else {
+      __pyx_t_3 = __pyx_t_6(__pyx_t_2);
+      if (unlikely(!__pyx_t_3)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1319, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_3);
+    }
+    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_3);
+    __pyx_t_3 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 1320:         ]
+
 1321:
+
 1322:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
+
+1323:         treatment_groups = sorted([x for x in list(set(treatment)) if x != self.control_name])
+
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_7 = PySet_New(__pyx_cur_scope->__pyx_v_treatment); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_4 = PySequence_List(__pyx_t_7); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_t_7 = __pyx_t_4; __Pyx_INCREF(__pyx_t_7); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  for (;;) {
+    if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_7)) break;
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    __pyx_t_4 = PyList_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_4); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1323, __pyx_L1_error)
+    #else
+    __pyx_t_4 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    #endif
+    __Pyx_XDECREF_SET(__pyx_v_x, __pyx_t_4);
+    __pyx_t_4 = 0;
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_3 = PyObject_RichCompare(__pyx_v_x, __pyx_t_4, Py_NE); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1323, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_11 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_11 < 0)) __PYX_ERR(0, 1323, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    if (__pyx_t_11) {
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_v_x))) __PYX_ERR(0, 1323, __pyx_L1_error)
+    }
+  }
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_t_1 = ((PyObject*)__pyx_t_2);
+  __pyx_t_2 = 0;
+  __pyx_t_12 = PyList_Sort(__pyx_t_1); if (unlikely(__pyx_t_12 == ((int)-1))) __PYX_ERR(0, 1323, __pyx_L1_error)
+  __pyx_v_treatment_groups = ((PyObject*)__pyx_t_1);
+  __pyx_t_1 = 0;
+
+1324:         self.classes_ = [self.control_name]
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1324, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_GIVEREF(__pyx_t_1);
+  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
+  __pyx_t_1 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 1324, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1325:         for tr in treatment_groups:
+
  if (unlikely(__pyx_v_treatment_groups == Py_None)) {
+    PyErr_SetString(PyExc_TypeError, "'NoneType' object is not iterable");
+    __PYX_ERR(0, 1325, __pyx_L1_error)
+  }
+  __pyx_t_2 = __pyx_v_treatment_groups; __Pyx_INCREF(__pyx_t_2); __pyx_t_5 = 0;
+  for (;;) {
+    if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_2)) break;
+    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+    __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1325, __pyx_L1_error)
+    #else
+    __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1325, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    #endif
+    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_1);
+    __pyx_t_1 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1326:             self.classes_.append(tr)
+
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1326, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_12 = __Pyx_PyObject_Append(__pyx_t_1, __pyx_v_tr); if (unlikely(__pyx_t_12 == ((int)-1))) __PYX_ERR(0, 1326, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+1327:         self.n_class = len(self.classes_)
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1327, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = PyObject_Length(__pyx_t_2); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1327, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1327, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_class, __pyx_t_2) < 0) __PYX_ERR(0, 1327, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
 1328:
+
+1329:         self.uplift_forest = (
+
  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest, __pyx_t_2) < 0) __PYX_ERR(0, 1329, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1330:             Parallel(n_jobs=self.n_jobs, prefer=self.joblib_prefer)
+
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_7 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_7, __pyx_n_s_n_jobs, __pyx_t_3) < 0) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_7, __pyx_n_s_prefer, __pyx_t_3) < 0) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_7); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
+1331:             (delayed(self.bootstrap)(X, treatment, y, tree) for tree in self.uplift_forest)
+
static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_genexpr(PyObject *__pyx_self) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *__pyx_cur_scope;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("genexpr", 0);
+  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr, __pyx_empty_tuple, NULL);
+  if (unlikely(!__pyx_cur_scope)) {
+    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *)Py_None);
+    __Pyx_INCREF(Py_None);
+    __PYX_ERR(0, 1331, __pyx_L1_error)
+  } else {
+    __Pyx_GOTREF(__pyx_cur_scope);
+  }
+  __pyx_cur_scope->__pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *) __pyx_self;
+  __Pyx_INCREF(((PyObject *)__pyx_cur_scope->__pyx_outer_scope));
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope);
+  {
+    __pyx_CoroutineObject *gen = __Pyx_Generator_New((__pyx_coroutine_body_t) __pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator, NULL, (PyObject *) __pyx_cur_scope, __pyx_n_s_genexpr, __pyx_n_s_UpliftRandomForestClassifier_fit, __pyx_n_s_causalml_inference_tree_uplift); if (unlikely(!gen)) __PYX_ERR(0, 1331, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_cur_scope);
+    __Pyx_RefNannyFinishContext();
+    return (PyObject *) gen;
+  }
+
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit.genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value) /* generator body */
+{
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("genexpr", 0);
+  __pyx_L3_first_run:;
+  if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1331, __pyx_L1_error) }
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
+    __pyx_t_2 = __pyx_t_1; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1331, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1331, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1331, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      }
+    } else {
+      __pyx_t_1 = __pyx_t_4(__pyx_t_2);
+      if (unlikely(!__pyx_t_1)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1331, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_1);
+    }
+    __Pyx_XGOTREF(__pyx_cur_scope->__pyx_v_tree);
+    __Pyx_XDECREF_SET(__pyx_cur_scope->__pyx_v_tree, __pyx_t_1);
+    __Pyx_GIVEREF(__pyx_t_1);
+    __pyx_t_1 = 0;
+    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_delayed); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1331, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1331, __pyx_L1_error) }
+    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_bootstrap); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __pyx_t_8 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
+      __pyx_t_8 = PyMethod_GET_SELF(__pyx_t_6);
+      if (likely(__pyx_t_8)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+        __Pyx_INCREF(__pyx_t_8);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_6, function);
+      }
+    }
+    __pyx_t_5 = (__pyx_t_8) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_8, __pyx_t_7) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_7);
+    __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
+    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1331, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X)) { __Pyx_RaiseClosureNameError("X"); __PYX_ERR(0, 1331, __pyx_L1_error) }
+    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment)) { __Pyx_RaiseClosureNameError("treatment"); __PYX_ERR(0, 1331, __pyx_L1_error) }
+    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y)) { __Pyx_RaiseClosureNameError("y"); __PYX_ERR(0, 1331, __pyx_L1_error) }
+    __pyx_t_6 = NULL;
+    __pyx_t_9 = 0;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_5);
+      if (likely(__pyx_t_6)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+        __Pyx_INCREF(__pyx_t_6);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_5, function);
+        __pyx_t_9 = 1;
+      }
+    }
+    #if CYTHON_FAST_PYCALL
+    if (PyFunction_Check(__pyx_t_5)) {
+      PyObject *__pyx_temp[5] = {__pyx_t_6, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y, __pyx_cur_scope->__pyx_v_tree};
+      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_9, 4+__pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_1);
+    } else
+    #endif
+    #if CYTHON_FAST_PYCCALL
+    if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
+      PyObject *__pyx_temp[5] = {__pyx_t_6, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y, __pyx_cur_scope->__pyx_v_tree};
+      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_9, 4+__pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+      __Pyx_GOTREF(__pyx_t_1);
+    } else
+    #endif
+    {
+      __pyx_t_7 = PyTuple_New(4+__pyx_t_9); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_7);
+      if (__pyx_t_6) {
+        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_6); __pyx_t_6 = NULL;
+      }
+      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
+      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
+      PyTuple_SET_ITEM(__pyx_t_7, 0+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
+      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
+      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
+      PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
+      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
+      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
+      PyTuple_SET_ITEM(__pyx_t_7, 2+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
+      __Pyx_INCREF(__pyx_cur_scope->__pyx_v_tree);
+      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_tree);
+      PyTuple_SET_ITEM(__pyx_t_7, 3+__pyx_t_9, __pyx_cur_scope->__pyx_v_tree);
+      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_7, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_r = __pyx_t_1;
+    __pyx_t_1 = 0;
+    __Pyx_XGIVEREF(__pyx_t_2);
+    __pyx_cur_scope->__pyx_t_0 = __pyx_t_2;
+    __pyx_cur_scope->__pyx_t_1 = __pyx_t_3;
+    __pyx_cur_scope->__pyx_t_2 = __pyx_t_4;
+    __Pyx_XGIVEREF(__pyx_r);
+    __Pyx_RefNannyFinishContext();
+    __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
+    /* return from generator, yielding value */
+    __pyx_generator->resume_label = 1;
+    return __pyx_r;
+    __pyx_L6_resume_from_yield:;
+    __pyx_t_2 = __pyx_cur_scope->__pyx_t_0;
+    __pyx_cur_scope->__pyx_t_0 = 0;
+    __Pyx_XGOTREF(__pyx_t_2);
+    __pyx_t_3 = __pyx_cur_scope->__pyx_t_1;
+    __pyx_t_4 = __pyx_cur_scope->__pyx_t_2;
+    if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  CYTHON_MAYBE_UNUSED_VAR(__pyx_cur_scope);
+
+  /* function exit code */
+  PyErr_SetNone(PyExc_StopIteration);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_AddTraceback("genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_r); __pyx_r = 0;
+  #if !CYTHON_USE_EXC_INFO_STACK
+  __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
+  #endif
+  __pyx_generator->resume_label = -1;
+  __Pyx_Coroutine_clear((PyObject*)__pyx_generator);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_t_7 = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_genexpr(((PyObject*)__pyx_cur_scope)); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __pyx_t_1 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
+    __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_3);
+    if (likely(__pyx_t_1)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
+      __Pyx_INCREF(__pyx_t_1);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_3, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_1, __pyx_t_7) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1331, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+/* … */
+struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr {
+  PyObject_HEAD
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *__pyx_outer_scope;
+  PyObject *__pyx_v_tree;
+  PyObject *__pyx_t_0;
+  Py_ssize_t __pyx_t_1;
+  PyObject *(*__pyx_t_2)(PyObject *);
+};
+
+
 1332:         )
+
 1333:
+
+1334:         all_importances = [tree.feature_importances_ for tree in self.uplift_forest]
+
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1334, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
+    __pyx_t_7 = __pyx_t_3; __Pyx_INCREF(__pyx_t_7); __pyx_t_5 = 0;
+    __pyx_t_6 = NULL;
+  } else {
+    __pyx_t_5 = -1; __pyx_t_7 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1334, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_7);
+    __pyx_t_6 = Py_TYPE(__pyx_t_7)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1334, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_6)) {
+      if (likely(PyList_CheckExact(__pyx_t_7))) {
+        if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_7)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1334, __pyx_L1_error)
+        #else
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        #endif
+      } else {
+        if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_7)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1334, __pyx_L1_error)
+        #else
+        __pyx_t_3 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_3);
+        #endif
+      }
+    } else {
+      __pyx_t_3 = __pyx_t_6(__pyx_t_7);
+      if (unlikely(!__pyx_t_3)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1334, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_3);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_tree, __pyx_t_3);
+    __pyx_t_3 = 0;
+    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_3);
+    if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_t_3))) __PYX_ERR(0, 1334, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_v_all_importances = ((PyObject*)__pyx_t_2);
+  __pyx_t_2 = 0;
+
+1335:         self.feature_importances_ = np.mean(all_importances, axis=0)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_mean); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_INCREF(__pyx_v_all_importances);
+  __Pyx_GIVEREF(__pyx_v_all_importances);
+  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_all_importances);
+  __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_axis, __pyx_int_0) < 0) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_1) < 0) __PYX_ERR(0, 1335, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
+1336:         self.feature_importances_ /= self.feature_importances_.sum()  # normalize to add to 1
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_t_2 = NULL;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_7))) {
+    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_7);
+    if (likely(__pyx_t_2)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_7);
+      __Pyx_INCREF(__pyx_t_2);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_7, function);
+    }
+  }
+  __pyx_t_3 = (__pyx_t_2) ? __Pyx_PyObject_CallOneArg(__pyx_t_7, __pyx_t_2) : __Pyx_PyObject_CallNoArg(__pyx_t_7);
+  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+  __pyx_t_7 = __Pyx_PyNumber_InPlaceDivide(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_7);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_7) < 0) __PYX_ERR(0, 1336, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
+
 1337:
+
+1338:     @staticmethod
+
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1338, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_bootstrap, __pyx_t_2) < 0) __PYX_ERR(0, 1339, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1339:     def bootstrap(X, treatment, y, tree):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap = {"bootstrap", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap, METH_VARARGS|METH_KEYWORDS, 0};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_treatment = 0;
+  PyObject *__pyx_v_y = 0;
+  PyObject *__pyx_v_tree = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("bootstrap (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_tree,0};
+    PyObject* values[4] = {0,0,0,0};
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+        CYTHON_FALLTHROUGH;
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 1); __PYX_ERR(0, 1339, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 2); __PYX_ERR(0, 1339, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  3:
+        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 3); __PYX_ERR(0, 1339, __pyx_L3_error)
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "bootstrap") < 0)) __PYX_ERR(0, 1339, __pyx_L3_error)
+      }
+    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
+      goto __pyx_L5_argtuple_error;
+    } else {
+      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
+    }
+    __pyx_v_X = values[0];
+    __pyx_v_treatment = values[1];
+    __pyx_v_y = values[2];
+    __pyx_v_tree = values[3];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1339, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_4bootstrap(__pyx_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_tree);
+  int __pyx_lineno = 0;
+  const char *__pyx_filename = NULL;
+  int __pyx_clineno = 0;
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_4bootstrap(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_tree) {
+  PyObject *__pyx_v_random_state = NULL;
+  PyObject *__pyx_v_bt_index = NULL;
+  PyObject *__pyx_v_x_train_bt = NULL;
+  PyObject *__pyx_v_y_train_bt = NULL;
+  PyObject *__pyx_v_treatment_train_bt = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("bootstrap", 0);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_random_state);
+  __Pyx_XDECREF(__pyx_v_bt_index);
+  __Pyx_XDECREF(__pyx_v_x_train_bt);
+  __Pyx_XDECREF(__pyx_v_y_train_bt);
+  __Pyx_XDECREF(__pyx_v_treatment_train_bt);
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__69 = PyTuple_Pack(9, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_random_state, __pyx_n_s_bt_index, __pyx_n_s_x_train_bt, __pyx_n_s_y_train_bt, __pyx_n_s_treatment_train_bt); if (unlikely(!__pyx_tuple__69)) __PYX_ERR(0, 1339, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__69);
+  __Pyx_GIVEREF(__pyx_tuple__69);
+/* … */
+  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftRandomForestClassifier_boo, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__70)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1339, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_codeobj__70 = (PyObject*)__Pyx_PyCode_New(4, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__69, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_bootstrap, 1339, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__70)) __PYX_ERR(0, 1339, __pyx_L1_error)
+
+1340:         random_state = check_random_state(tree.random_state)
+
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1340, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1340, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_4 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_4)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_4);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+    }
+  }
+  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1340, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_random_state = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1341:         bt_index = random_state.choice(len(X), len(X))
+
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_random_state, __pyx_n_s_choice); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1341, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __pyx_t_5 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1341, __pyx_L1_error)
+  __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1341, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __pyx_t_5 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1341, __pyx_L1_error)
+  __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1341, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_6 = NULL;
+  __pyx_t_7 = 0;
+  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
+    __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
+    if (likely(__pyx_t_6)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
+      __Pyx_INCREF(__pyx_t_6);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_2, function);
+      __pyx_t_7 = 1;
+    }
+  }
+  #if CYTHON_FAST_PYCALL
+  if (PyFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, __pyx_t_4};
+    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_7, 2+__pyx_t_7); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  } else
+  #endif
+  #if CYTHON_FAST_PYCCALL
+  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
+    PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, __pyx_t_4};
+    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_7, 2+__pyx_t_7); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
+    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  } else
+  #endif
+  {
+    __pyx_t_8 = PyTuple_New(2+__pyx_t_7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1341, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    if (__pyx_t_6) {
+      __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_t_6); __pyx_t_6 = NULL;
+    }
+    __Pyx_GIVEREF(__pyx_t_3);
+    PyTuple_SET_ITEM(__pyx_t_8, 0+__pyx_t_7, __pyx_t_3);
+    __Pyx_GIVEREF(__pyx_t_4);
+    PyTuple_SET_ITEM(__pyx_t_8, 1+__pyx_t_7, __pyx_t_4);
+    __pyx_t_3 = 0;
+    __pyx_t_4 = 0;
+    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_8, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __pyx_v_bt_index = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1342:         x_train_bt = X[bt_index]
+
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1342, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_x_train_bt = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1343:         y_train_bt = y[bt_index]
+
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1343, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_y_train_bt = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1344:         treatment_train_bt = treatment[bt_index]
+
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1344, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_v_treatment_train_bt = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1345:         tree.fit(X=x_train_bt, treatment=treatment_train_bt, y=y_train_bt)
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_fit); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1345, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1345, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_x_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_treatment, __pyx_v_treatment_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
+  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_y, __pyx_v_y_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
+  __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1345, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+1346:         return tree
+
  __Pyx_XDECREF(__pyx_r);
+  __Pyx_INCREF(__pyx_v_tree);
+  __pyx_r = __pyx_v_tree;
+  goto __pyx_L0;
+
 1347:
+
+1348:     @ignore_warnings(category=FutureWarning)
+
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1348, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1348, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_category, __pyx_builtin_FutureWarning) < 0) __PYX_ERR(0, 1348, __pyx_L1_error)
+  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_empty_tuple, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1348, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_6);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+
+1349:     def predict(self, X, full_output=False):
+
/* Python wrapper */
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
+static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict[] = "\n        Returns the recommended treatment group and predicted optimal\n        probability conditional on using the recommended treatment group.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        full_output : bool, optional (default=False)\n            Whether the UpliftTree algorithm returns upliftScores, pred_nodes\n            alongside the recommended treatment group and p_hat in the treatment group.\n\n        Returns\n        -------\n        y_pred_list : ndarray, shape = (num_samples, num_treatments])\n            An ndarray containing the predicted treatment effect of each treatment group for each sample\n\n        df_res : DataFrame, shape = [num_samples, (num_treatments * 2 + 3)]\n            If `full_output` is `True`, a DataFrame containing the predicted outcome of each treatment and\n            control group, the treatment effect of each treatment group, the treatment group with the\n            highest treatment effect, and the maximum treatment effect for each sample.\n\n        ";
+static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict};
+static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
+  PyObject *__pyx_v_self = 0;
+  PyObject *__pyx_v_X = 0;
+  PyObject *__pyx_v_full_output = 0;
+  PyObject *__pyx_r = 0;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
+  {
+    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_full_output,0};
+    PyObject* values[3] = {0,0,0};
+    values[2] = ((PyObject *)((PyObject *)Py_False));
+    if (unlikely(__pyx_kwds)) {
+      Py_ssize_t kw_args;
+      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
+      switch (pos_args) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        CYTHON_FALLTHROUGH;
+        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        CYTHON_FALLTHROUGH;
+        case  0: break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+      kw_args = PyDict_Size(__pyx_kwds);
+      switch (pos_args) {
+        case  0:
+        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
+        else goto __pyx_L5_argtuple_error;
+        CYTHON_FALLTHROUGH;
+        case  1:
+        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
+        else {
+          __Pyx_RaiseArgtupleInvalid("predict", 0, 2, 3, 1); __PYX_ERR(0, 1349, __pyx_L3_error)
+        }
+        CYTHON_FALLTHROUGH;
+        case  2:
+        if (kw_args > 0) {
+          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_full_output);
+          if (value) { values[2] = value; kw_args--; }
+        }
+      }
+      if (unlikely(kw_args > 0)) {
+        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 1349, __pyx_L3_error)
+      }
+    } else {
+      switch (PyTuple_GET_SIZE(__pyx_args)) {
+        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
+        CYTHON_FALLTHROUGH;
+        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
+        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
+        break;
+        default: goto __pyx_L5_argtuple_error;
+      }
+    }
+    __pyx_v_self = values[0];
+    __pyx_v_X = values[1];
+    __pyx_v_full_output = values[2];
+  }
+  goto __pyx_L4_argument_unpacking_done;
+  __pyx_L5_argtuple_error:;
+  __Pyx_RaiseArgtupleInvalid("predict", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1349, __pyx_L3_error)
+  __pyx_L3_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __Pyx_RefNannyFinishContext();
+  return NULL;
+  __pyx_L4_argument_unpacking_done:;
+  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_full_output);
+
+  /* function exit code */
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value); /* proto */
+/* … */
+static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_full_output) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *__pyx_cur_scope;
+  PyObject *__pyx_v_y_pred_ensemble = NULL;
+  PyObject *__pyx_v_df_res = NULL;
+  PyObject *__pyx_v_delta_cols = NULL;
+  PyObject *__pyx_v_i_tr = NULL;
+  PyObject *__pyx_v_treatment_group = NULL;
+  PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1 = 0;
+  PyObject *__pyx_v_tree = NULL;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("predict", 0);
+  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict, __pyx_empty_tuple, NULL);
+  if (unlikely(!__pyx_cur_scope)) {
+    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *)Py_None);
+    __Pyx_INCREF(Py_None);
+    __PYX_ERR(0, 1349, __pyx_L1_error)
+  } else {
+    __Pyx_GOTREF(__pyx_cur_scope);
+  }
+  __pyx_cur_scope->__pyx_v_self = __pyx_v_self;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_self);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_self);
+  __pyx_cur_scope->__pyx_v_X = __pyx_v_X;
+  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_X);
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_X);
+/* … */
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_4);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_8);
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_v_y_pred_ensemble);
+  __Pyx_XDECREF(__pyx_v_df_res);
+  __Pyx_XDECREF(__pyx_v_delta_cols);
+  __Pyx_XDECREF(__pyx_v_i_tr);
+  __Pyx_XDECREF(__pyx_v_treatment_group);
+  __Pyx_XDECREF(__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1);
+  __Pyx_XDECREF(__pyx_v_tree);
+  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+  __pyx_tuple__71 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_full_output, __pyx_n_s_y_pred_ensemble, __pyx_n_s_df_res, __pyx_n_s_delta_cols, __pyx_n_s_i_tr, __pyx_n_s_treatment_group, __pyx_n_s_genexpr, __pyx_n_s_genexpr, __pyx_n_s_tree); if (unlikely(!__pyx_tuple__71)) __PYX_ERR(0, 1349, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_tuple__71);
+  __Pyx_GIVEREF(__pyx_tuple__71);
+  __pyx_codeobj__72 = (PyObject*)__Pyx_PyCode_New(3, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__71, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_predict, 1349, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__72)) __PYX_ERR(0, 1349, __pyx_L1_error)
+/* … */
+  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict, 0, __pyx_n_s_UpliftRandomForestClassifier_pre_2, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__72)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1349, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_3);
+  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_3, __pyx_tuple__73);
+  __pyx_t_5 = NULL;
+  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
+    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
+    if (likely(__pyx_t_5)) {
+      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
+      __Pyx_INCREF(__pyx_t_5);
+      __Pyx_INCREF(function);
+      __Pyx_DECREF_SET(__pyx_t_6, function);
+    }
+  }
+  __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_5, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_3);
+  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
+  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1348, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_predict, __pyx_t_2) < 0) __PYX_ERR(0, 1349, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+/* … */
+struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict {
+  PyObject_HEAD
+  PyObject *__pyx_v_X;
+  PyObject *__pyx_v_self;
+};
+
+
 1350:         '''
+
 1351:         Returns the recommended treatment group and predicted optimal
+
 1352:         probability conditional on using the recommended treatment group.
+
 1353:
+
 1354:         Args
+
 1355:         ----
+
 1356:         X : ndarray, shape = [num_samples, num_features]
+
 1357:             An ndarray of the covariates used to train the uplift model.
+
 1358:
+
 1359:         full_output : bool, optional (default=False)
+
 1360:             Whether the UpliftTree algorithm returns upliftScores, pred_nodes
+
 1361:             alongside the recommended treatment group and p_hat in the treatment group.
+
 1362:
+
 1363:         Returns
+
 1364:         -------
+
 1365:         y_pred_list : ndarray, shape = (num_samples, num_treatments])
+
 1366:             An ndarray containing the predicted treatment effect of each treatment group for each sample
+
 1367:
+
 1368:         df_res : DataFrame, shape = [num_samples, (num_treatments * 2 + 3)]
+
 1369:             If `full_output` is `True`, a DataFrame containing the predicted outcome of each treatment and
+
 1370:             control group, the treatment effect of each treatment group, the treatment group with the
+
 1371:             highest treatment effect, and the maximum treatment effect for each sample.
+
 1372:
+
 1373:         '''
+
 1374:         # Make predictions with all trees and take the average
+
 1375:
+
+1376:         if self.n_jobs != 1:
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1376, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __pyx_t_2 = __Pyx_PyInt_NeObjC(__pyx_t_1, __pyx_int_1, 1, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1376, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_2);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1376, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  if (__pyx_t_3) {
+/* … */
+    goto __pyx_L3;
+  }
+
+1377:             y_pred_ensemble = sum(
+
    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1377, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+
+1378:                 Parallel(n_jobs=self.n_jobs, prefer=self.joblib_prefer)
+
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_n_jobs, __pyx_t_5) < 0) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_prefer, __pyx_t_5) < 0) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
+1379:                 (delayed(tree.predict)(X=X) for tree in self.uplift_forest)
+
static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_genexpr(PyObject *__pyx_self) {
+  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *__pyx_cur_scope;
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("genexpr", 0);
+  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr, __pyx_empty_tuple, NULL);
+  if (unlikely(!__pyx_cur_scope)) {
+    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *)Py_None);
+    __Pyx_INCREF(Py_None);
+    __PYX_ERR(0, 1379, __pyx_L1_error)
+  } else {
+    __Pyx_GOTREF(__pyx_cur_scope);
+  }
+  __pyx_cur_scope->__pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *) __pyx_self;
+  __Pyx_INCREF(((PyObject *)__pyx_cur_scope->__pyx_outer_scope));
+  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope);
+  {
+    __pyx_CoroutineObject *gen = __Pyx_Generator_New((__pyx_coroutine_body_t) __pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1, NULL, (PyObject *) __pyx_cur_scope, __pyx_n_s_genexpr, __pyx_n_s_UpliftRandomForestClassifier_pre, __pyx_n_s_causalml_inference_tree_uplift); if (unlikely(!gen)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_cur_scope);
+    __Pyx_RefNannyFinishContext();
+    return (PyObject *) gen;
+  }
+
+  /* function exit code */
+  __pyx_L1_error:;
+  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict.genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_r = NULL;
+  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
+  __Pyx_XGIVEREF(__pyx_r);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+
+static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value) /* generator body */
+{
+  PyObject *__pyx_r = NULL;
+  __Pyx_RefNannyDeclarations
+  __Pyx_RefNannySetupContext("genexpr", 0);
+  __pyx_L3_first_run:;
+  if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1379, __pyx_L1_error) }
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
+    __pyx_t_2 = __pyx_t_1; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
+    __pyx_t_4 = NULL;
+  } else {
+    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_4)) {
+      if (likely(PyList_CheckExact(__pyx_t_2))) {
+        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1379, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      } else {
+        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1379, __pyx_L1_error)
+        #else
+        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_1);
+        #endif
+      }
+    } else {
+      __pyx_t_1 = __pyx_t_4(__pyx_t_2);
+      if (unlikely(!__pyx_t_1)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1379, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_1);
+    }
+    __Pyx_XGOTREF(__pyx_cur_scope->__pyx_v_tree);
+    __Pyx_XDECREF_SET(__pyx_cur_scope->__pyx_v_tree, __pyx_t_1);
+    __Pyx_GIVEREF(__pyx_t_1);
+    __pyx_t_1 = 0;
+    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_delayed); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_tree, __pyx_n_s_predict); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __pyx_t_7 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_5);
+      if (likely(__pyx_t_7)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+        __Pyx_INCREF(__pyx_t_7);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_5, function);
+      }
+    }
+    __pyx_t_1 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_7, __pyx_t_6) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_6);
+    __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
+    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
+    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X)) { __Pyx_RaiseClosureNameError("X"); __PYX_ERR(0, 1379, __pyx_L1_error) }
+    if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X) < 0) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_6);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_r = __pyx_t_6;
+    __pyx_t_6 = 0;
+    __Pyx_XGIVEREF(__pyx_t_2);
+    __pyx_cur_scope->__pyx_t_0 = __pyx_t_2;
+    __pyx_cur_scope->__pyx_t_1 = __pyx_t_3;
+    __pyx_cur_scope->__pyx_t_2 = __pyx_t_4;
+    __Pyx_XGIVEREF(__pyx_r);
+    __Pyx_RefNannyFinishContext();
+    __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
+    /* return from generator, yielding value */
+    __pyx_generator->resume_label = 1;
+    return __pyx_r;
+    __pyx_L6_resume_from_yield:;
+    __pyx_t_2 = __pyx_cur_scope->__pyx_t_0;
+    __pyx_cur_scope->__pyx_t_0 = 0;
+    __Pyx_XGOTREF(__pyx_t_2);
+    __pyx_t_3 = __pyx_cur_scope->__pyx_t_1;
+    __pyx_t_4 = __pyx_cur_scope->__pyx_t_2;
+    if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1379, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+  CYTHON_MAYBE_UNUSED_VAR(__pyx_cur_scope);
+
+  /* function exit code */
+  PyErr_SetNone(PyExc_StopIteration);
+  goto __pyx_L0;
+  __pyx_L1_error:;
+  __Pyx_XDECREF(__pyx_t_1);
+  __Pyx_XDECREF(__pyx_t_2);
+  __Pyx_XDECREF(__pyx_t_5);
+  __Pyx_XDECREF(__pyx_t_6);
+  __Pyx_XDECREF(__pyx_t_7);
+  __Pyx_AddTraceback("genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
+  __pyx_L0:;
+  __Pyx_XDECREF(__pyx_r); __pyx_r = 0;
+  #if !CYTHON_USE_EXC_INFO_STACK
+  __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
+  #endif
+  __pyx_generator->resume_label = -1;
+  __Pyx_Coroutine_clear((PyObject*)__pyx_generator);
+  __Pyx_RefNannyFinishContext();
+  return __pyx_r;
+}
+/* … */
+    __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_genexpr(((PyObject*)__pyx_cur_scope)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_1 = NULL;
+    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
+      __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_5);
+      if (likely(__pyx_t_1)) {
+        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
+        __Pyx_INCREF(__pyx_t_1);
+        __Pyx_INCREF(function);
+        __Pyx_DECREF_SET(__pyx_t_5, function);
+      }
+    }
+    __pyx_t_2 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_1, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_4);
+    __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1379, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+
+1380:             ) / len(self.uplift_forest)
+
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1380, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_6 = PyObject_Length(__pyx_t_2); if (unlikely(__pyx_t_6 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1380, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_t_2 = PyInt_FromSsize_t(__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1380, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    __pyx_t_4 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1380, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    __pyx_v_y_pred_ensemble = __pyx_t_4;
+    __pyx_t_4 = 0;
+
 1381:         else:
+
+1382:             y_pred_ensemble = sum([tree.predict(X=X) for tree in self.uplift_forest]) / len(self.uplift_forest)
+
  /*else*/ {
+    __pyx_t_4 = PyList_New(0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_2);
+    if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
+      __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_6 = 0;
+      __pyx_t_7 = NULL;
+    } else {
+      __pyx_t_6 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_5);
+      __pyx_t_7 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    }
+    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+    for (;;) {
+      if (likely(!__pyx_t_7)) {
+        if (likely(PyList_CheckExact(__pyx_t_5))) {
+          if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_5)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1382, __pyx_L1_error)
+          #else
+          __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_2);
+          #endif
+        } else {
+          if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
+          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+          __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1382, __pyx_L1_error)
+          #else
+          __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
+          __Pyx_GOTREF(__pyx_t_2);
+          #endif
+        }
+      } else {
+        __pyx_t_2 = __pyx_t_7(__pyx_t_5);
+        if (unlikely(!__pyx_t_2)) {
+          PyObject* exc_type = PyErr_Occurred();
+          if (exc_type) {
+            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+            else __PYX_ERR(0, 1382, __pyx_L1_error)
+          }
+          break;
+        }
+        __Pyx_GOTREF(__pyx_t_2);
+      }
+      __Pyx_XDECREF_SET(__pyx_v_tree, __pyx_t_2);
+      __pyx_t_2 = 0;
+      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_2);
+      __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_1);
+      if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_X, __pyx_cur_scope->__pyx_v_X) < 0) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __Pyx_GOTREF(__pyx_t_8);
+      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
+      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+      if (unlikely(__Pyx_ListComp_Append(__pyx_t_4, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 1382, __pyx_L1_error)
+      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+    }
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_6 = PyObject_Length(__pyx_t_4); if (unlikely(__pyx_t_6 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_8 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1382, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_v_y_pred_ensemble = __pyx_t_8;
+    __pyx_t_8 = 0;
+  }
+  __pyx_L3:;
+
 1383:
+
 1384:         # Summarize results into dataframe
+
+1385:         df_res = pd.DataFrame(y_pred_ensemble, columns=self.classes_)
+
  __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_pd); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_DataFrame); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  __pyx_t_8 = PyTuple_New(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __Pyx_INCREF(__pyx_v_y_pred_ensemble);
+  __Pyx_GIVEREF(__pyx_v_y_pred_ensemble);
+  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_y_pred_ensemble);
+  __pyx_t_5 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_columns, __pyx_t_1) < 0) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1385, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_v_df_res = __pyx_t_1;
+  __pyx_t_1 = 0;
+
+1386:         df_res['recommended_treatment'] = df_res.apply(np.argmax, axis=1)
+
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_df_res, __pyx_n_s_apply); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_argmax); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_GIVEREF(__pyx_t_8);
+  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_8);
+  __pyx_t_8 = 0;
+  __pyx_t_8 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_n_s_recommended_treatment, __pyx_t_4) < 0)) __PYX_ERR(0, 1386, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
 1387:
+
 1388:         # Calculate delta
+
+1389:         delta_cols = [f'delta_{treatment_group}' for treatment_group in self.classes_[1:]]
+
  __pyx_t_4 = PyList_New(0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1389, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1389, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __pyx_t_5 = __Pyx_PyObject_GetSlice(__pyx_t_8, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_5);
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_5)) || PyTuple_CheckExact(__pyx_t_5)) {
+    __pyx_t_8 = __pyx_t_5; __Pyx_INCREF(__pyx_t_8); __pyx_t_6 = 0;
+    __pyx_t_7 = NULL;
+  } else {
+    __pyx_t_6 = -1; __pyx_t_8 = PyObject_GetIter(__pyx_t_5); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __pyx_t_7 = Py_TYPE(__pyx_t_8)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1389, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_7)) {
+      if (likely(PyList_CheckExact(__pyx_t_8))) {
+        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_8)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1389, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      } else {
+        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_8)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1389, __pyx_L1_error)
+        #else
+        __pyx_t_5 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_5);
+        #endif
+      }
+    } else {
+      __pyx_t_5 = __pyx_t_7(__pyx_t_8);
+      if (unlikely(!__pyx_t_5)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1389, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_5);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_5);
+    __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_FormatSimple(__pyx_v_treatment_group, __pyx_empty_unicode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_n_u_delta, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1389, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(__Pyx_ListComp_Append(__pyx_t_4, (PyObject*)__pyx_t_1))) __PYX_ERR(0, 1389, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+  }
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  __pyx_v_delta_cols = ((PyObject*)__pyx_t_4);
+  __pyx_t_4 = 0;
+
+1390:         for i_tr in range(1, self.n_class):
+
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __pyx_t_8 = PyTuple_New(2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1390, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __Pyx_INCREF(__pyx_int_1);
+  __Pyx_GIVEREF(__pyx_int_1);
+  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_int_1);
+  __Pyx_GIVEREF(__pyx_t_4);
+  PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_t_4);
+  __pyx_t_4 = 0;
+  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_8, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
+    __pyx_t_8 = __pyx_t_4; __Pyx_INCREF(__pyx_t_8); __pyx_t_6 = 0;
+    __pyx_t_7 = NULL;
+  } else {
+    __pyx_t_6 = -1; __pyx_t_8 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1390, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __pyx_t_7 = Py_TYPE(__pyx_t_8)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1390, __pyx_L1_error)
+  }
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  for (;;) {
+    if (likely(!__pyx_t_7)) {
+      if (likely(PyList_CheckExact(__pyx_t_8))) {
+        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_8)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1390, __pyx_L1_error)
+        #else
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        #endif
+      } else {
+        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_8)) break;
+        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
+        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1390, __pyx_L1_error)
+        #else
+        __pyx_t_4 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
+        __Pyx_GOTREF(__pyx_t_4);
+        #endif
+      }
+    } else {
+      __pyx_t_4 = __pyx_t_7(__pyx_t_8);
+      if (unlikely(!__pyx_t_4)) {
+        PyObject* exc_type = PyErr_Occurred();
+        if (exc_type) {
+          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
+          else __PYX_ERR(0, 1390, __pyx_L1_error)
+        }
+        break;
+      }
+      __Pyx_GOTREF(__pyx_t_4);
+    }
+    __Pyx_XDECREF_SET(__pyx_v_i_tr, __pyx_t_4);
+    __pyx_t_4 = 0;
+/* … */
+  }
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+
+1391:             treatment_group = self.classes_[i_tr]
+
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1391, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_t_4, __pyx_v_i_tr); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1391, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_1);
+    __pyx_t_1 = 0;
+
+1392:             df_res[f'delta_{treatment_group}'] = df_res[treatment_group] - df_res[self.control_name]
+
    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_treatment_group); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+    __pyx_t_4 = PyNumber_Subtract(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_4);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    __pyx_t_5 = __Pyx_PyObject_FormatSimple(__pyx_v_treatment_group, __pyx_empty_unicode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_5);
+    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_n_u_delta, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
+    if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_t_1, __pyx_t_4) < 0)) __PYX_ERR(0, 1392, __pyx_L1_error)
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+
 1393:
+
+1394:         df_res['max_delta'] = df_res[delta_cols].max(axis=1)
+
  __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_delta_cols); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_max); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_4);
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  __pyx_t_8 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_8);
+  if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __Pyx_GOTREF(__pyx_t_1);
+  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
+  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
+  if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_n_s_max_delta, __pyx_t_1) < 0)) __PYX_ERR(0, 1394, __pyx_L1_error)
+  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+
 1395:
+
+1396:         if full_output:
+
  __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_v_full_output); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1396, __pyx_L1_error)
+  if (__pyx_t_3) {
+/* … */
+  }
+
+1397:             return df_res
+
    __Pyx_XDECREF(__pyx_r);
+    __Pyx_INCREF(__pyx_v_df_res);
+    __pyx_r = __pyx_v_df_res;
+    goto __pyx_L0;
+
 1398:         else:
+
+1399:             return df_res[delta_cols].values
+
  /*else*/ {
+    __Pyx_XDECREF(__pyx_r);
+    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_delta_cols); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1399, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_1);
+    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_values); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1399, __pyx_L1_error)
+    __Pyx_GOTREF(__pyx_t_8);
+    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
+    __pyx_r = __pyx_t_8;
+    __pyx_t_8 = 0;
+    goto __pyx_L0;
+  }
+
diff --git a/requirements-tf.txt b/requirements-tf.txt index aff9d594..acd535d5 100644 --- a/requirements-tf.txt +++ b/requirements-tf.txt @@ -11,7 +11,7 @@ Cython>=0.28.0 xgboost pydotplus tqdm -shap<0.38.1 +shap dill lightgbm pygam diff --git a/requirements.txt b/requirements.txt index a0fa8e39..9679ebfd 100644 --- a/requirements.txt +++ b/requirements.txt @@ -11,7 +11,7 @@ Cython>=0.28.0 xgboost pydotplus tqdm -shap<0.38.1 +shap dill lightgbm pygam From 722e7c2e399428719077dd1d2d7524f64cc06f45 Mon Sep 17 00:00:00 2001 From: jpan2 Date: Fri, 11 Mar 2022 10:43:09 -0800 Subject: [PATCH 2/2] remove html files --- causalml/inference/tree/causaltree.html | 3588 ------ causalml/inference/tree/uplift.html | 13953 ---------------------- 2 files changed, 17541 deletions(-) delete mode 100644 causalml/inference/tree/causaltree.html delete mode 100644 causalml/inference/tree/uplift.html diff --git a/causalml/inference/tree/causaltree.html b/causalml/inference/tree/causaltree.html deleted file mode 100644 index a07ba4e7..00000000 --- a/causalml/inference/tree/causaltree.html +++ /dev/null @@ -1,3588 +0,0 @@ - - - - - - Cython: causaltree.pyx - - - -

Generated by Cython 0.29.24

-

- Yellow lines hint at Python interaction.
- Click on a line that starts with a "+" to see the C code that Cython generated for it. -

-

Raw output: causaltree.c

-
+001: # cython: cdivision=True
-
  __pyx_t_2 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_2) < 0) __PYX_ERR(0, 1, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 002: # cython: boundscheck=False
-
 003: # cython: wraparound=False
-
 004:
-
 005:
-
+006: import logging
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_logging, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 6, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_logging, __pyx_t_1) < 0) __PYX_ERR(0, 6, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+007: import numbers
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numbers, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 7, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_numbers, __pyx_t_1) < 0) __PYX_ERR(0, 7, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+008: import numpy as np
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 8, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 8, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+009: import pandas as pd
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_pandas, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 9, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_pd, __pyx_t_1) < 0) __PYX_ERR(0, 9, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 010:
-
+011: from math import ceil
-
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 11, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_ceil);
-  __Pyx_GIVEREF(__pyx_n_s_ceil);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_ceil);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_math, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 11, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_ceil); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 11, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_ceil, __pyx_t_1) < 0) __PYX_ERR(0, 11, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+012: from scipy.sparse import issparse
-
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 12, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s_issparse);
-  __Pyx_GIVEREF(__pyx_n_s_issparse);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_issparse);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_scipy_sparse, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 12, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_issparse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 12, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_issparse, __pyx_t_2) < 0) __PYX_ERR(0, 12, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+013: from scipy.stats import norm
-
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 13, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_norm);
-  __Pyx_GIVEREF(__pyx_n_s_norm);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_norm);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_scipy_stats, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 13, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_norm); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 13, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_norm, __pyx_t_1) < 0) __PYX_ERR(0, 13, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 014: from sklearn.tree._criterion cimport RegressionCriterion
-
 015: from sklearn.tree._criterion cimport SIZE_t, DOUBLE_t
-
+016: from sklearn.tree._splitter import BestSplitter
-
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 16, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s_BestSplitter);
-  __Pyx_GIVEREF(__pyx_n_s_BestSplitter);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_BestSplitter);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn_tree__splitter, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 16, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_BestSplitter); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 16, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_BestSplitter, __pyx_t_2) < 0) __PYX_ERR(0, 16, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+017: from sklearn.tree._tree import DepthFirstTreeBuilder, DOUBLE, DTYPE, Tree
-
  __pyx_t_1 = PyList_New(4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_DepthFirstTreeBuilder);
-  __Pyx_GIVEREF(__pyx_n_s_DepthFirstTreeBuilder);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_DepthFirstTreeBuilder);
-  __Pyx_INCREF(__pyx_n_s_DOUBLE);
-  __Pyx_GIVEREF(__pyx_n_s_DOUBLE);
-  PyList_SET_ITEM(__pyx_t_1, 1, __pyx_n_s_DOUBLE);
-  __Pyx_INCREF(__pyx_n_s_DTYPE);
-  __Pyx_GIVEREF(__pyx_n_s_DTYPE);
-  PyList_SET_ITEM(__pyx_t_1, 2, __pyx_n_s_DTYPE);
-  __Pyx_INCREF(__pyx_n_s_Tree);
-  __Pyx_GIVEREF(__pyx_n_s_Tree);
-  PyList_SET_ITEM(__pyx_t_1, 3, __pyx_n_s_Tree);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_tree__tree, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DepthFirstTreeBuilder); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DepthFirstTreeBuilder, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DOUBLE, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DTYPE, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_Tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Tree, __pyx_t_1) < 0) __PYX_ERR(0, 17, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+018: from sklearn.utils import check_array, check_random_state
-
  __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s_check_array);
-  __Pyx_GIVEREF(__pyx_n_s_check_array);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_check_array);
-  __Pyx_INCREF(__pyx_n_s_check_random_state);
-  __Pyx_GIVEREF(__pyx_n_s_check_random_state);
-  PyList_SET_ITEM(__pyx_t_2, 1, __pyx_n_s_check_random_state);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn_utils, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_check_array); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_array, __pyx_t_2) < 0) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_random_state, __pyx_t_2) < 0) __PYX_ERR(0, 18, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 019:
-
+020: from causalml.inference.meta.utils import check_treatment_vector
-
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_check_treatment_vector);
-  __Pyx_GIVEREF(__pyx_n_s_check_treatment_vector);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_check_treatment_vector);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_causalml_inference_meta_utils, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_treatment_vector); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_treatment_vector, __pyx_t_1) < 0) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 021:
-
+022: logger = logging.getLogger('causalml')
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_logging); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_getLogger); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_tuple__25, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_logger, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-/* … */
-  __pyx_tuple__25 = PyTuple_Pack(1, __pyx_n_s_causalml); if (unlikely(!__pyx_tuple__25)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__25);
-  __Pyx_GIVEREF(__pyx_tuple__25);
-
 023:
-
 024:
-
+025: cdef class CausalMSE(RegressionCriterion):
-
struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE {
-  struct __pyx_obj_7sklearn_4tree_10_criterion_RegressionCriterion __pyx_base;
-};
-/* … */
-struct __pyx_vtabstruct_8causalml_9inference_4tree_10causaltree_CausalMSE {
-  struct __pyx_vtabstruct_7sklearn_4tree_10_criterion_RegressionCriterion __pyx_base;
-};
-static struct __pyx_vtabstruct_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_vtabptr_8causalml_9inference_4tree_10causaltree_CausalMSE;
-
-
 026:     """Causal Tree mean squared error impurity criterion.
-
 027:
-
 028:         CausalTreeMSE = right_effect + left_effect
-
 029:
-
 030:         where,
-
 031:
-
 032:         effect = alpha * tau^2 - (1 - alpha) * (1 + train_to_est_ratio) * (VAR_tr / p + VAR_cont / (1 - p))
-
 033:     """
-
 034:
-
+035:     cdef void node_value(self, double* dest) nogil:
-
static void __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_node_value(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self, double *__pyx_v_dest) {
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
-  double __pyx_v_node_ct;
-  double __pyx_v_node_tr;
-  double __pyx_v_node_ct_sum;
-  double __pyx_v_node_tr_sum;
-  double __pyx_v_one_over_eps;
-/* … */
-  /* function exit code */
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.node_value", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
-  __pyx_L0:;
-}
-
 036:         """Compute the node value of samples[start:end] into dest."""
-
 037:
-
+038:         cdef SIZE_t start = self.start
-
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.start;
-  __pyx_v_start = __pyx_t_1;
-
+039:         cdef SIZE_t end = self.end
-
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.end;
-  __pyx_v_end = __pyx_t_1;
-
 040:
-
 041:         cdef SIZE_t i
-
 042:         cdef SIZE_t p
-
 043:         cdef DOUBLE_t is_treated
-
 044:         cdef DOUBLE_t y_ik
-
 045:
-
+046:         cdef SIZE_t* samples = self.samples
-
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.samples;
-  __pyx_v_samples = __pyx_t_2;
-
+047:         cdef DOUBLE_t* sample_weight = self.sample_weight
-
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
-  __pyx_v_sample_weight = __pyx_t_3;
-
 048:
-
+049:         cdef double node_ct = 0.0
-
  __pyx_v_node_ct = 0.0;
-
+050:         cdef double node_tr = 0.0
-
  __pyx_v_node_tr = 0.0;
-
+051:         cdef double node_ct_sum = 0.0
-
  __pyx_v_node_ct_sum = 0.0;
-
+052:         cdef double node_tr_sum = 0.0
-
  __pyx_v_node_tr_sum = 0.0;
-
+053:         cdef double one_over_eps = 1e5
-
  __pyx_v_one_over_eps = 1e5;
-
 054:
-
+055:         for p in range(start, end):
-
  __pyx_t_1 = __pyx_v_end;
-  __pyx_t_4 = __pyx_t_1;
-  for (__pyx_t_5 = __pyx_v_start; __pyx_t_5 < __pyx_t_4; __pyx_t_5+=1) {
-    __pyx_v_p = __pyx_t_5;
-
+056:             i = samples[p]
-
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
-
 057:
-
+058:             if sample_weight != NULL:
-
    __pyx_t_6 = ((__pyx_v_sample_weight != NULL) != 0);
-    if (__pyx_t_6) {
-/* … */
-    }
-
 059:                 # the weights of 1 and 1 + eps are used for treatment and control respectively
-
+060:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
-
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
-
 061:
-
 062:             # assume that there is only one output (k = 0)
-
+063:             y_ik = self.y[i, 0]
-
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 63, __pyx_L1_error)}
-    __pyx_t_7 = __pyx_v_i;
-    __pyx_t_8 = 0;
-    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_7 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_8)) )));
-
 064:
-
+065:             node_tr += is_treated
-
    __pyx_v_node_tr = (__pyx_v_node_tr + __pyx_v_is_treated);
-
+066:             node_ct += 1. - is_treated
-
    __pyx_v_node_ct = (__pyx_v_node_ct + (1. - __pyx_v_is_treated));
-
+067:             node_tr_sum += y_ik * is_treated
-
    __pyx_v_node_tr_sum = (__pyx_v_node_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
-
+068:             node_ct_sum += y_ik * (1. - is_treated)
-
    __pyx_v_node_ct_sum = (__pyx_v_node_ct_sum + (__pyx_v_y_ik * (1. - __pyx_v_is_treated)));
-  }
-
 069:
-
 070:         # save the average of treatment effects within a node as a value for the node
-
+071:         dest[0] = node_tr_sum / node_tr - node_ct_sum / node_ct
-
  (__pyx_v_dest[0]) = ((__pyx_v_node_tr_sum / __pyx_v_node_tr) - (__pyx_v_node_ct_sum / __pyx_v_node_ct));
-
 072:
-
+073:     cdef double node_impurity(self) nogil:
-
static double __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_node_impurity(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self) {
-  CYTHON_UNUSED double *__pyx_v_sum_total;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
-  double __pyx_v_node_tr;
-  double __pyx_v_node_ct;
-  double __pyx_v_node_sum;
-  double __pyx_v_node_tr_sum;
-  double __pyx_v_node_sq_sum;
-  double __pyx_v_node_tr_sq_sum;
-  double __pyx_v_tr_var;
-  double __pyx_v_ct_var;
-  double __pyx_v_one_over_eps;
-  double __pyx_v_node_tau;
-  double __pyx_r;
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.node_impurity", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
-  __pyx_r = 0;
-  __pyx_L0:;
-  return __pyx_r;
-}
-
 074:         """Evaluate the impurity of the current node, i.e. the impurity of
-
 075:            samples[start:end]."""
-
 076:
-
+077:         cdef double* sum_total = self.sum_total
-
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.sum_total;
-  __pyx_v_sum_total = __pyx_t_1;
-
 078:         cdef double impurity
-
+079:         cdef SIZE_t start = self.start
-
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.start;
-  __pyx_v_start = __pyx_t_2;
-
+080:         cdef SIZE_t end = self.end
-
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.end;
-  __pyx_v_end = __pyx_t_2;
-
 081:
-
 082:         cdef SIZE_t i
-
 083:         cdef SIZE_t p
-
 084:         cdef DOUBLE_t is_treated
-
 085:         cdef DOUBLE_t y_ik
-
 086:
-
+087:         cdef SIZE_t* samples = self.samples
-
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.samples;
-  __pyx_v_samples = __pyx_t_3;
-
+088:         cdef DOUBLE_t* sample_weight = self.sample_weight
-
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
-  __pyx_v_sample_weight = __pyx_t_4;
-
 089:
-
+090:         cdef double node_tr = 0.0
-
  __pyx_v_node_tr = 0.0;
-
+091:         cdef double node_ct = 0.0
-
  __pyx_v_node_ct = 0.0;
-
+092:         cdef double node_sum = self.sum_total[0]
-
  __pyx_v_node_sum = (__pyx_v_self->__pyx_base.__pyx_base.sum_total[0]);
-
+093:         cdef double node_tr_sum = 0.0
-
  __pyx_v_node_tr_sum = 0.0;
-
+094:         cdef double node_sq_sum = 0.0
-
  __pyx_v_node_sq_sum = 0.0;
-
+095:         cdef double node_tr_sq_sum = 0.0
-
  __pyx_v_node_tr_sq_sum = 0.0;
-
 096:         cdef double tr_var
-
 097:         cdef double ct_var
-
+098:         cdef double one_over_eps = 1e5
-
  __pyx_v_one_over_eps = 1e5;
-
 099:
-
+100:         for p in range(start, end):
-
  __pyx_t_2 = __pyx_v_end;
-  __pyx_t_5 = __pyx_t_2;
-  for (__pyx_t_6 = __pyx_v_start; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
-    __pyx_v_p = __pyx_t_6;
-
+101:             i = samples[p]
-
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
-
 102:
-
+103:             if sample_weight != NULL:
-
    __pyx_t_7 = ((__pyx_v_sample_weight != NULL) != 0);
-    if (__pyx_t_7) {
-/* … */
-    }
-
 104:                 # the weights of 1 and 1 + eps are used for treatment and control respectively
-
+105:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
-
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
-
 106:
-
 107:             # assume that there is only one output (k = 0)
-
+108:             y_ik = self.y[i, 0]
-
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 108, __pyx_L1_error)}
-    __pyx_t_8 = __pyx_v_i;
-    __pyx_t_9 = 0;
-    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_8 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_9)) )));
-
 109:
-
+110:             node_tr += is_treated
-
    __pyx_v_node_tr = (__pyx_v_node_tr + __pyx_v_is_treated);
-
+111:             node_ct += (1. - is_treated)
-
    __pyx_v_node_ct = (__pyx_v_node_ct + (1. - __pyx_v_is_treated));
-
+112:             node_tr_sum += y_ik * is_treated
-
    __pyx_v_node_tr_sum = (__pyx_v_node_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
-
+113:             node_sq_sum += y_ik * y_ik
-
    __pyx_v_node_sq_sum = (__pyx_v_node_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
-
+114:             node_tr_sq_sum += y_ik * y_ik * is_treated
-
    __pyx_v_node_tr_sq_sum = (__pyx_v_node_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
-  }
-
 115:
-
+116:         node_tau = node_tr_sum / node_tr - (node_sum - node_tr_sum) / node_ct
-
  __pyx_v_node_tau = ((__pyx_v_node_tr_sum / __pyx_v_node_tr) - ((__pyx_v_node_sum - __pyx_v_node_tr_sum) / __pyx_v_node_ct));
-
+117:         tr_var = node_tr_sq_sum / node_tr - node_tr_sum * node_tr_sum / (node_tr * node_tr)
-
  __pyx_v_tr_var = ((__pyx_v_node_tr_sq_sum / __pyx_v_node_tr) - ((__pyx_v_node_tr_sum * __pyx_v_node_tr_sum) / (__pyx_v_node_tr * __pyx_v_node_tr)));
-
+118:         ct_var = ((node_sq_sum - node_tr_sq_sum) / node_ct -
-
  __pyx_v_ct_var = (((__pyx_v_node_sq_sum - __pyx_v_node_tr_sq_sum) / __pyx_v_node_ct) - (((__pyx_v_node_sum - __pyx_v_node_tr_sum) * (__pyx_v_node_sum - __pyx_v_node_tr_sum)) / (__pyx_v_node_ct * __pyx_v_node_ct)));
-
 119:                   (node_sum - node_tr_sum) * (node_sum - node_tr_sum) / (node_ct * node_ct))
-
 120:
-
+121:         return  (tr_var / node_tr + ct_var / node_ct) - node_tau * node_tau
-
  __pyx_r = (((__pyx_v_tr_var / __pyx_v_node_tr) + (__pyx_v_ct_var / __pyx_v_node_ct)) - (__pyx_v_node_tau * __pyx_v_node_tau));
-  goto __pyx_L0;
-
 122:
-
 123:
-
+124:     cdef void children_impurity(self, double* impurity_left, double* impurity_right) nogil:
-
static void __pyx_f_8causalml_9inference_4tree_10causaltree_9CausalMSE_children_impurity(struct __pyx_obj_8causalml_9inference_4tree_10causaltree_CausalMSE *__pyx_v_self, double *__pyx_v_impurity_left, double *__pyx_v_impurity_right) {
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t *__pyx_v_sample_weight;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t *__pyx_v_samples;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_start;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_pos;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_end;
-  double *__pyx_v_sum_left;
-  double *__pyx_v_sum_right;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_i;
-  __pyx_t_7sklearn_4tree_5_tree_SIZE_t __pyx_v_p;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_is_treated;
-  __pyx_t_7sklearn_4tree_5_tree_DOUBLE_t __pyx_v_y_ik;
-  double __pyx_v_right_tr;
-  double __pyx_v_right_ct;
-  double __pyx_v_right_sum;
-  double __pyx_v_right_tr_sum;
-  double __pyx_v_right_sq_sum;
-  double __pyx_v_right_tr_sq_sum;
-  double __pyx_v_right_tr_var;
-  double __pyx_v_right_ct_var;
-  double __pyx_v_left_tr;
-  double __pyx_v_left_ct;
-  double __pyx_v_left_sum;
-  double __pyx_v_left_tr_sum;
-  double __pyx_v_left_sq_sum;
-  double __pyx_v_left_tr_sq_sum;
-  double __pyx_v_left_tr_var;
-  double __pyx_v_left_ct_var;
-  double __pyx_v_one_over_eps;
-  double __pyx_v_right_tau;
-  double __pyx_v_left_tau;
-/* … */
-  /* function exit code */
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_WriteUnraisable("causalml.inference.tree.causaltree.CausalMSE.children_impurity", __pyx_clineno, __pyx_lineno, __pyx_filename, 1, 1);
-  __pyx_L0:;
-}
-
 125:         """Evaluate the impurity in children nodes, i.e. the impurity of the
-
 126:            left child (samples[start:pos]) and the impurity the right child
-
 127:            (samples[pos:end])."""
-
 128:
-
+129:         cdef DOUBLE_t* sample_weight = self.sample_weight
-
  __pyx_t_1 = __pyx_v_self->__pyx_base.__pyx_base.sample_weight;
-  __pyx_v_sample_weight = __pyx_t_1;
-
+130:         cdef SIZE_t* samples = self.samples
-
  __pyx_t_2 = __pyx_v_self->__pyx_base.__pyx_base.samples;
-  __pyx_v_samples = __pyx_t_2;
-
+131:         cdef SIZE_t start = self.start
-
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.start;
-  __pyx_v_start = __pyx_t_3;
-
+132:         cdef SIZE_t pos = self.pos
-
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.pos;
-  __pyx_v_pos = __pyx_t_3;
-
+133:         cdef SIZE_t end = self.end
-
  __pyx_t_3 = __pyx_v_self->__pyx_base.__pyx_base.end;
-  __pyx_v_end = __pyx_t_3;
-
 134:
-
+135:         cdef double* sum_left = self.sum_left
-
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sum_left;
-  __pyx_v_sum_left = __pyx_t_4;
-
+136:         cdef double* sum_right = self.sum_right
-
  __pyx_t_4 = __pyx_v_self->__pyx_base.__pyx_base.sum_right;
-  __pyx_v_sum_right = __pyx_t_4;
-
 137:
-
 138:         cdef SIZE_t i
-
 139:         cdef SIZE_t p
-
 140:         cdef DOUBLE_t is_treated
-
 141:         cdef DOUBLE_t y_ik
-
 142:
-
+143:         cdef double right_tr = 0.0
-
  __pyx_v_right_tr = 0.0;
-
+144:         cdef double right_ct = 0.0
-
  __pyx_v_right_ct = 0.0;
-
+145:         cdef double right_sum = 0.0
-
  __pyx_v_right_sum = 0.0;
-
+146:         cdef double right_tr_sum = 0.0
-
  __pyx_v_right_tr_sum = 0.0;
-
+147:         cdef double right_sq_sum = 0.0
-
  __pyx_v_right_sq_sum = 0.0;
-
+148:         cdef double right_tr_sq_sum = 0.0
-
  __pyx_v_right_tr_sq_sum = 0.0;
-
 149:         cdef double right_tr_var
-
 150:         cdef double right_ct_var
-
 151:
-
+152:         cdef double left_tr = 0.0
-
  __pyx_v_left_tr = 0.0;
-
+153:         cdef double left_ct = 0.0
-
  __pyx_v_left_ct = 0.0;
-
+154:         cdef double left_sum = 0.0
-
  __pyx_v_left_sum = 0.0;
-
+155:         cdef double left_tr_sum = 0.0
-
  __pyx_v_left_tr_sum = 0.0;
-
+156:         cdef double left_sq_sum = 0.0
-
  __pyx_v_left_sq_sum = 0.0;
-
+157:         cdef double left_tr_sq_sum = 0.0
-
  __pyx_v_left_tr_sq_sum = 0.0;
-
 158:         cdef double left_tr_var
-
 159:         cdef double left_ct_var
-
 160:
-
+161:         cdef double one_over_eps = 1e5
-
  __pyx_v_one_over_eps = 1e5;
-
 162:
-
+163:         for p in range(start, end):
-
  __pyx_t_3 = __pyx_v_end;
-  __pyx_t_5 = __pyx_t_3;
-  for (__pyx_t_6 = __pyx_v_start; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
-    __pyx_v_p = __pyx_t_6;
-
+164:             i = samples[p]
-
    __pyx_v_i = (__pyx_v_samples[__pyx_v_p]);
-
 165:
-
+166:             if sample_weight != NULL:
-
    __pyx_t_7 = ((__pyx_v_sample_weight != NULL) != 0);
-    if (__pyx_t_7) {
-/* … */
-    }
-
 167:                 # the weights of 1 and 1 + eps are used for control and treatment respectively
-
+168:                 is_treated = (sample_weight[i] - 1.0) * one_over_eps
-
      __pyx_v_is_treated = (((__pyx_v_sample_weight[__pyx_v_i]) - 1.0) * __pyx_v_one_over_eps);
-
 169:
-
 170:             # assume that there is only one output (k = 0)
-
+171:             y_ik = self.y[i, 0]
-
    if (unlikely(!__pyx_v_self->__pyx_base.__pyx_base.y.memview)) {PyErr_SetString(PyExc_AttributeError,"Memoryview is not initialized");__PYX_ERR(0, 171, __pyx_L1_error)}
-    __pyx_t_8 = __pyx_v_i;
-    __pyx_t_9 = 0;
-    __pyx_v_y_ik = (*((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=1 */ ((char *) (((__pyx_t_7sklearn_4tree_5_tree_DOUBLE_t const  *) ( /* dim=0 */ (__pyx_v_self->__pyx_base.__pyx_base.y.data + __pyx_t_8 * __pyx_v_self->__pyx_base.__pyx_base.y.strides[0]) )) + __pyx_t_9)) )));
-
 172:
-
+173:             if p < pos:
-
    __pyx_t_7 = ((__pyx_v_p < __pyx_v_pos) != 0);
-    if (__pyx_t_7) {
-/* … */
-      goto __pyx_L6;
-    }
-
+174:                 left_tr += is_treated
-
      __pyx_v_left_tr = (__pyx_v_left_tr + __pyx_v_is_treated);
-
+175:                 left_ct += 1. - is_treated
-
      __pyx_v_left_ct = (__pyx_v_left_ct + (1. - __pyx_v_is_treated));
-
+176:                 left_sum += y_ik
-
      __pyx_v_left_sum = (__pyx_v_left_sum + __pyx_v_y_ik);
-
+177:                 left_tr_sum += y_ik * is_treated
-
      __pyx_v_left_tr_sum = (__pyx_v_left_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
-
+178:                 left_sq_sum += y_ik * y_ik
-
      __pyx_v_left_sq_sum = (__pyx_v_left_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
-
+179:                 left_tr_sq_sum += y_ik * y_ik * is_treated
-
      __pyx_v_left_tr_sq_sum = (__pyx_v_left_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
-
 180:             else:
-
+181:                 right_tr += is_treated
-
    /*else*/ {
-      __pyx_v_right_tr = (__pyx_v_right_tr + __pyx_v_is_treated);
-
+182:                 right_ct += 1. - is_treated
-
      __pyx_v_right_ct = (__pyx_v_right_ct + (1. - __pyx_v_is_treated));
-
+183:                 right_sum += y_ik
-
      __pyx_v_right_sum = (__pyx_v_right_sum + __pyx_v_y_ik);
-
+184:                 right_tr_sum += y_ik * is_treated
-
      __pyx_v_right_tr_sum = (__pyx_v_right_tr_sum + (__pyx_v_y_ik * __pyx_v_is_treated));
-
+185:                 right_sq_sum += y_ik * y_ik
-
      __pyx_v_right_sq_sum = (__pyx_v_right_sq_sum + (__pyx_v_y_ik * __pyx_v_y_ik));
-
+186:                 right_tr_sq_sum += y_ik * y_ik * is_treated
-
      __pyx_v_right_tr_sq_sum = (__pyx_v_right_tr_sq_sum + ((__pyx_v_y_ik * __pyx_v_y_ik) * __pyx_v_is_treated));
-    }
-    __pyx_L6:;
-  }
-
 187:
-
+188:         right_tau = right_tr_sum / right_tr - (sum_right[0] - right_tr_sum) / right_ct
-
  __pyx_v_right_tau = ((__pyx_v_right_tr_sum / __pyx_v_right_tr) - (((__pyx_v_sum_right[0]) - __pyx_v_right_tr_sum) / __pyx_v_right_ct));
-
+189:         right_tr_var = right_tr_sq_sum / right_tr - right_tr_sum * right_tr_sum / (right_tr * right_tr)
-
  __pyx_v_right_tr_var = ((__pyx_v_right_tr_sq_sum / __pyx_v_right_tr) - ((__pyx_v_right_tr_sum * __pyx_v_right_tr_sum) / (__pyx_v_right_tr * __pyx_v_right_tr)));
-
+190:         right_ct_var = ((right_sq_sum - right_tr_sq_sum) / right_ct -
-
  __pyx_v_right_ct_var = (((__pyx_v_right_sq_sum - __pyx_v_right_tr_sq_sum) / __pyx_v_right_ct) - (((__pyx_v_right_sum - __pyx_v_right_tr_sum) * (__pyx_v_right_sum - __pyx_v_right_tr_sum)) / (__pyx_v_right_ct * __pyx_v_right_ct)));
-
 191:                          (right_sum - right_tr_sum) * (right_sum - right_tr_sum) / (right_ct * right_ct))
-
 192:
-
+193:         left_tau = left_tr_sum / left_tr - (sum_left[0] - left_tr_sum) / left_ct
-
  __pyx_v_left_tau = ((__pyx_v_left_tr_sum / __pyx_v_left_tr) - (((__pyx_v_sum_left[0]) - __pyx_v_left_tr_sum) / __pyx_v_left_ct));
-
+194:         left_tr_var = left_tr_sq_sum / left_tr - left_tr_sum * left_tr_sum / (left_tr * left_tr)
-
  __pyx_v_left_tr_var = ((__pyx_v_left_tr_sq_sum / __pyx_v_left_tr) - ((__pyx_v_left_tr_sum * __pyx_v_left_tr_sum) / (__pyx_v_left_tr * __pyx_v_left_tr)));
-
+195:         left_ct_var = ((left_sq_sum - left_tr_sq_sum) / left_ct -
-
  __pyx_v_left_ct_var = (((__pyx_v_left_sq_sum - __pyx_v_left_tr_sq_sum) / __pyx_v_left_ct) - (((__pyx_v_left_sum - __pyx_v_left_tr_sum) * (__pyx_v_left_sum - __pyx_v_left_tr_sum)) / (__pyx_v_left_ct * __pyx_v_left_ct)));
-
 196:                         (left_sum - left_tr_sum) * (left_sum - left_tr_sum) / (left_ct * left_ct))
-
 197:
-
+198:         impurity_left[0] = (left_tr_var / left_tr + left_ct_var / left_ct) - left_tau * left_tau
-
  (__pyx_v_impurity_left[0]) = (((__pyx_v_left_tr_var / __pyx_v_left_tr) + (__pyx_v_left_ct_var / __pyx_v_left_ct)) - (__pyx_v_left_tau * __pyx_v_left_tau));
-
+199:         impurity_right[0] = (right_tr_var / right_tr + right_ct_var / right_ct) - right_tau * right_tau
-
  (__pyx_v_impurity_right[0]) = (((__pyx_v_right_tr_var / __pyx_v_right_tr) + (__pyx_v_right_ct_var / __pyx_v_right_ct)) - (__pyx_v_right_tau * __pyx_v_right_tau));
-
 200:
-
 201:
-
+202: class CausalTreeRegressor:
-
  __pyx_t_2 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_CausalTreeRegressor, __pyx_n_s_CausalTreeRegressor, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_kp_s_A_Causal_Tree_regressor_class_Th); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 202, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-/* … */
-  __pyx_t_1 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_CausalTreeRegressor, __pyx_empty_tuple, __pyx_t_2, NULL, 0, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 202, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_CausalTreeRegressor, __pyx_t_1) < 0) __PYX_ERR(0, 202, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 203:     """A Causal Tree regressor class.
-
 204:
-
 205:     The Causal Tree is a decision tree regressor with a split criteria for treatment effects instead of
-
 206:     outputs.
-
 207:
-
 208:     Details are available at Athey and Imbens (2015) (https://arxiv.org/abs/1504.01132)
-
 209:     """
-
+210:     def __init__(self, ate_alpha=.05, control_name=0, max_depth=None,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__[] = "Initialize a Causal Tree\n\n        Args:\n            ate_alpha (float, optional): the confidence level alpha of the ATE estimate\n            control_name (str or int, optional): name of control group\n            max_depth (int, optional): the maximum depth of tree\n            min_samples_leaf (int, optional): the minimum number of samples in leaves\n            random_state (int or np.RandomState, optional): a random seed or a random state\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_ate_alpha = 0;
-  PyObject *__pyx_v_control_name = 0;
-  PyObject *__pyx_v_max_depth = 0;
-  PyObject *__pyx_v_min_samples_leaf = 0;
-  PyObject *__pyx_v_random_state = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_ate_alpha,&__pyx_n_s_control_name,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_random_state,0};
-    PyObject* values[6] = {0,0,0,0,0,0};
-    values[1] = ((PyObject *)((PyObject*)__pyx_float__05));
-    values[2] = ((PyObject *)((PyObject *)__pyx_int_0));
-    values[3] = ((PyObject *)((PyObject *)Py_None));
-    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
-/* … */
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_ate_alpha, PyObject *__pyx_v_control_name, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_random_state) {
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__", 0);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__26 = PyTuple_Pack(6, __pyx_n_s_self, __pyx_n_s_ate_alpha, __pyx_n_s_control_name, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_random_state); if (unlikely(!__pyx_tuple__26)) __PYX_ERR(0, 210, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__26);
-  __Pyx_GIVEREF(__pyx_tuple__26);
-  __pyx_codeobj__27 = (PyObject*)__Pyx_PyCode_New(6, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__26, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_init, 210, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__27)) __PYX_ERR(0, 210, __pyx_L1_error)
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_1__init__, 0, __pyx_n_s_CausalTreeRegressor___init, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__27)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 210, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__28);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_init, __pyx_t_1) < 0) __PYX_ERR(0, 210, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_tuple__28 = PyTuple_Pack(5, ((PyObject*)__pyx_float__05), ((PyObject *)__pyx_int_0), ((PyObject *)Py_None), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__28)) __PYX_ERR(0, 210, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__28);
-  __Pyx_GIVEREF(__pyx_tuple__28);
-
+211:                  min_samples_leaf=100, random_state=None):
-
    values[5] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_ate_alpha);
-          if (value) { values[1] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name);
-          if (value) { values[2] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
-          if (value) { values[3] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
-          if (value) { values[5] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 210, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_ate_alpha = values[1];
-    __pyx_v_control_name = values[2];
-    __pyx_v_max_depth = values[3];
-    __pyx_v_min_samples_leaf = values[4];
-    __pyx_v_random_state = values[5];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 0, 1, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 210, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor___init__(__pyx_self, __pyx_v_self, __pyx_v_ate_alpha, __pyx_v_control_name, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_random_state);
-
 212:         """Initialize a Causal Tree
-
 213:
-
 214:         Args:
-
 215:             ate_alpha (float, optional): the confidence level alpha of the ATE estimate
-
 216:             control_name (str or int, optional): name of control group
-
 217:             max_depth (int, optional): the maximum depth of tree
-
 218:             min_samples_leaf (int, optional): the minimum number of samples in leaves
-
 219:             random_state (int or np.RandomState, optional): a random seed or a random state
-
 220:         """
-
+221:         self.ate_alpha = ate_alpha
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha, __pyx_v_ate_alpha) < 0) __PYX_ERR(0, 221, __pyx_L1_error)
-
+222:         self.control_name = control_name
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 222, __pyx_L1_error)
-
+223:         self.max_depth = max_depth
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 223, __pyx_L1_error)
-
+224:         self.min_samples_leaf = min_samples_leaf
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 224, __pyx_L1_error)
-
+225:         self.random_state = random_state
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 225, __pyx_L1_error)
-
 226:
-
+227:         self._classes = {}
-
  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 227, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_1) < 0) __PYX_ERR(0, 227, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+228:         self.tree = None
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_tree, Py_None) < 0) __PYX_ERR(0, 228, __pyx_L1_error)
-
 229:
-
+230:         self.eps = 1e-5
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_eps, __pyx_float_1eneg_5) < 0) __PYX_ERR(0, 230, __pyx_L1_error)
-
 231:
-
+232:     def fit(self, X, treatment, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit[] = "Fit the Causal Tree model\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n\n        Returns:\n            self (CausalTree object)\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 232, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 232, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 232, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 232, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 232, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_is_treatment = NULL;
-  PyObject *__pyx_v_w = NULL;
-  PyObject *__pyx_v_t_groups = NULL;
-  PyObject *__pyx_v_random_state = NULL;
-  PyObject *__pyx_v_n_samples = NULL;
-  PyObject *__pyx_v_n_features = NULL;
-  PyObject *__pyx_v_n_outputs = NULL;
-  PyObject *__pyx_v_min_samples_leaf = NULL;
-  PyObject *__pyx_v_max_depth = NULL;
-  PyObject *__pyx_v_splitter = NULL;
-  PyObject *__pyx_v_builder = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit", 0);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_INCREF(__pyx_v_y);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_XDECREF(__pyx_t_12);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_is_treatment);
-  __Pyx_XDECREF(__pyx_v_w);
-  __Pyx_XDECREF(__pyx_v_t_groups);
-  __Pyx_XDECREF(__pyx_v_random_state);
-  __Pyx_XDECREF(__pyx_v_n_samples);
-  __Pyx_XDECREF(__pyx_v_n_features);
-  __Pyx_XDECREF(__pyx_v_n_outputs);
-  __Pyx_XDECREF(__pyx_v_min_samples_leaf);
-  __Pyx_XDECREF(__pyx_v_max_depth);
-  __Pyx_XDECREF(__pyx_v_splitter);
-  __Pyx_XDECREF(__pyx_v_builder);
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XDECREF(__pyx_v_y);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__29 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_is_treatment, __pyx_n_s_w, __pyx_n_s_t_groups, __pyx_n_s_random_state, __pyx_n_s_n_samples, __pyx_n_s_n_features, __pyx_n_s_n_outputs, __pyx_n_s_min_samples_leaf, __pyx_n_s_max_depth, __pyx_n_s_splitter, __pyx_n_s_builder); if (unlikely(!__pyx_tuple__29)) __PYX_ERR(0, 232, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__29);
-  __Pyx_GIVEREF(__pyx_tuple__29);
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_3fit, 0, __pyx_n_s_CausalTreeRegressor_fit, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__30)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 232, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_fit, __pyx_t_1) < 0) __PYX_ERR(0, 232, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_codeobj__30 = (PyObject*)__Pyx_PyCode_New(4, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__29, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_fit, 232, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__30)) __PYX_ERR(0, 232, __pyx_L1_error)
-
 233:         """Fit the Causal Tree model
-
 234:
-
 235:         Args:
-
 236:             X (np.matrix): a feature matrix
-
 237:             treatment (np.array): a treatment vector
-
 238:             y (np.array): an outcome vector
-
 239:
-
 240:         Returns:
-
 241:             self (CausalTree object)
-
 242:         """
-
+243:         check_treatment_vector(treatment, self.control_name)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_treatment_vector); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 243, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 243, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = NULL;
-  __pyx_t_5 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_5 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment, __pyx_t_3};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment, __pyx_t_3};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  } else
-  #endif
-  {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 243, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    if (__pyx_t_4) {
-      __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_4); __pyx_t_4 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_treatment);
-    __Pyx_GIVEREF(__pyx_v_treatment);
-    PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_5, __pyx_v_treatment);
-    __Pyx_GIVEREF(__pyx_t_3);
-    PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 243, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+244:         is_treatment = treatment != self.control_name
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 244, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_t_1, Py_NE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 244, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_is_treatment = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+245:         w = is_treatment.astype(int)
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_is_treatment, __pyx_n_s_astype); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 245, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_6 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_6)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_6);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_6, ((PyObject *)(&PyInt_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_1, ((PyObject *)(&PyInt_Type)));
-  __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 245, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_w = __pyx_t_2;
-  __pyx_t_2 = 0;
-
 246:
-
+247:         t_groups = np.unique(treatment[is_treatment])
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 247, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_unique); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 247, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_is_treatment); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 247, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_6);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_6, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 247, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_v_t_groups = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+248:         self._classes[t_groups[0]] = 0
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 248, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_t_groups, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 248, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_t_6, __pyx_int_0) < 0)) __PYX_ERR(0, 248, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-
 249:
-
 250:         # input checking replicated from BaseDecisionTree.fit()
-
+251:         random_state = check_random_state(self.random_state)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 251, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 251, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_6 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 251, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_random_state = __pyx_t_6;
-  __pyx_t_6 = 0;
-
+252:         X = check_array(X, dtype=DTYPE, accept_sparse="csc")
-
  __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_check_array); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_v_X);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
-  __pyx_t_1 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_3) < 0) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_accept_sparse, __pyx_n_s_csc) < 0) __PYX_ERR(0, 252, __pyx_L1_error)
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_3);
-  __pyx_t_3 = 0;
-
+253:         y = check_array(y, ensure_2d=False, dtype=None)
-
  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_check_array); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 253, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 253, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_v_y);
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 253, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_ensure_2d, Py_False) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_dtype, Py_None) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
-  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 253, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
-  __pyx_t_6 = 0;
-
+254:         if issparse(X):
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_issparse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_1)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_1);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_6 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_1, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
-  __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (__pyx_t_7) {
-/* … */
-  }
-
+255:             X.sort_indices()
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_sort_indices); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 255, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = NULL;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-      __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_2);
-      if (likely(__pyx_t_1)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-        __Pyx_INCREF(__pyx_t_1);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_2, function);
-      }
-    }
-    __pyx_t_6 = (__pyx_t_1) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_1) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-    __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
-    if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 255, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-
 256:
-
+257:             if X.indices.dtype != np.intc or X.indptr.dtype != np.intc:
-
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_indices); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_dtype); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_intc); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = PyObject_RichCompare(__pyx_t_2, __pyx_t_1, Py_NE); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    if (!__pyx_t_8) {
-    } else {
-      __pyx_t_7 = __pyx_t_8;
-      goto __pyx_L5_bool_binop_done;
-    }
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_indptr); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_dtype); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_intc); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = PyObject_RichCompare(__pyx_t_1, __pyx_t_2, Py_NE); __Pyx_XGOTREF(__pyx_t_6); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_6); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 257, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_7 = __pyx_t_8;
-    __pyx_L5_bool_binop_done:;
-    if (unlikely(__pyx_t_7)) {
-/* … */
-    }
-
+258:                 raise ValueError("No support for np.int64 index based "
-
      __pyx_t_6 = __Pyx_PyObject_Call(__pyx_builtin_ValueError, __pyx_tuple__3, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 258, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_6);
-      __Pyx_Raise(__pyx_t_6, 0, 0, 0);
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __PYX_ERR(0, 258, __pyx_L1_error)
-/* … */
-  __pyx_tuple__3 = PyTuple_Pack(1, __pyx_kp_s_No_support_for_np_int64_index_ba); if (unlikely(!__pyx_tuple__3)) __PYX_ERR(0, 258, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__3);
-  __Pyx_GIVEREF(__pyx_tuple__3);
-
 259:                                  "sparse matrices")
-
 260:
-
+261:         y = np.atleast_1d(y)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 261, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_atleast_1d); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 261, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_6 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_2, __pyx_v_y) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_y);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 261, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
-  __pyx_t_6 = 0;
-
+262:         if y.ndim == 1:
-
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_ndim); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 262, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_1 = __Pyx_PyInt_EqObjC(__pyx_t_6, __pyx_int_1, 1, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 262, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 262, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (__pyx_t_7) {
-/* … */
-  }
-
+263:             y = np.reshape(y, (-1, 1))
-
    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 263, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_reshape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 263, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = NULL;
-    __pyx_t_5 = 0;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
-      if (likely(__pyx_t_6)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-        __Pyx_INCREF(__pyx_t_6);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_2, function);
-        __pyx_t_5 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_2)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_y, __pyx_tuple__4};
-      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_1);
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_v_y, __pyx_tuple__4};
-      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_1);
-    } else
-    #endif
-    {
-      __pyx_t_3 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 263, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      if (__pyx_t_6) {
-        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_6); __pyx_t_6 = NULL;
-      }
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_5, __pyx_v_y);
-      __Pyx_INCREF(__pyx_tuple__4);
-      __Pyx_GIVEREF(__pyx_tuple__4);
-      PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_5, __pyx_tuple__4);
-      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_3, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 263, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_1);
-    __pyx_t_1 = 0;
-/* … */
-  __pyx_tuple__4 = PyTuple_Pack(2, __pyx_int_neg_1, __pyx_int_1); if (unlikely(!__pyx_tuple__4)) __PYX_ERR(0, 263, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__4);
-  __Pyx_GIVEREF(__pyx_tuple__4);
-
+264:         if getattr(y, "dtype", None) != DOUBLE or not y.flags.contiguous:
-
  __pyx_t_1 = __Pyx_GetAttr3(__pyx_v_y, __pyx_n_s_dtype, Py_None); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = PyObject_RichCompare(__pyx_t_1, __pyx_t_2, Py_NE); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (!__pyx_t_8) {
-  } else {
-    __pyx_t_7 = __pyx_t_8;
-    goto __pyx_L9_bool_binop_done;
-  }
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_flags); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_contiguous); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 264, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_9 = ((!__pyx_t_8) != 0);
-  __pyx_t_7 = __pyx_t_9;
-  __pyx_L9_bool_binop_done:;
-  if (__pyx_t_7) {
-/* … */
-  }
-
+265:             y = np.ascontiguousarray(y, dtype=DOUBLE)
-
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_ascontiguousarray); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_DOUBLE); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_6) < 0) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 265, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_6);
-    __pyx_t_6 = 0;
-
+266:         n_samples, n_features = X.shape
-
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 266, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
-    PyObject* sequence = __pyx_t_6;
-    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-    if (unlikely(size != 2)) {
-      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-      __PYX_ERR(0, 266, __pyx_L1_error)
-    }
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    if (likely(PyTuple_CheckExact(sequence))) {
-      __pyx_t_1 = PyTuple_GET_ITEM(sequence, 0); 
-      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 1); 
-    } else {
-      __pyx_t_1 = PyList_GET_ITEM(sequence, 0); 
-      __pyx_t_2 = PyList_GET_ITEM(sequence, 1); 
-    }
-    __Pyx_INCREF(__pyx_t_1);
-    __Pyx_INCREF(__pyx_t_2);
-    #else
-    __pyx_t_1 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 266, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 266, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    #endif
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  } else {
-    Py_ssize_t index = -1;
-    __pyx_t_3 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 266, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_10 = Py_TYPE(__pyx_t_3)->tp_iternext;
-    index = 0; __pyx_t_1 = __pyx_t_10(__pyx_t_3); if (unlikely(!__pyx_t_1)) goto __pyx_L11_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_1);
-    index = 1; __pyx_t_2 = __pyx_t_10(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L11_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_3), 2) < 0) __PYX_ERR(0, 266, __pyx_L1_error)
-    __pyx_t_10 = NULL;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    goto __pyx_L12_unpacking_done;
-    __pyx_L11_unpacking_failed:;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_10 = NULL;
-    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-    __PYX_ERR(0, 266, __pyx_L1_error)
-    __pyx_L12_unpacking_done:;
-  }
-  __pyx_v_n_samples = __pyx_t_1;
-  __pyx_t_1 = 0;
-  __pyx_v_n_features = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+267:         n_outputs = y.shape[1]
-
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_y, __pyx_n_s_shape); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 267, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_6, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 267, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_v_n_outputs = __pyx_t_2;
-  __pyx_t_2 = 0;
-
 268:
-
+269:         if isinstance(self.min_samples_leaf, numbers.Integral):
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 269, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_numbers); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 269, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_Integral); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 269, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_t_7 = PyObject_IsInstance(__pyx_t_2, __pyx_t_1); if (unlikely(__pyx_t_7 == ((int)-1))) __PYX_ERR(0, 269, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_9 = (__pyx_t_7 != 0);
-  if (__pyx_t_9) {
-/* … */
-    goto __pyx_L13;
-  }
-
+270:             if not 1 <= self.min_samples_leaf:
-
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 270, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = PyObject_RichCompare(__pyx_int_1, __pyx_t_1, Py_LE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 270, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 270, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_7 = ((!__pyx_t_9) != 0);
-    if (unlikely(__pyx_t_7)) {
-/* … */
-    }
-
+271:                 raise ValueError("min_samples_leaf must be at least 1 "
-
      __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_ValueError, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 271, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __Pyx_Raise(__pyx_t_2, 0, 0, 0);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __PYX_ERR(0, 271, __pyx_L1_error)
-
 272:                                  "or in (0, 0.5], got %s"
-
+273:                                  % self.min_samples_leaf)
-
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 273, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_1 = __Pyx_PyString_FormatSafe(__pyx_kp_s_min_samples_leaf_must_be_at_leas, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 273, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+274:             min_samples_leaf = self.min_samples_leaf
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 274, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_v_min_samples_leaf = __pyx_t_2;
-    __pyx_t_2 = 0;
-
 275:         else:  # float
-
+276:             if not 0. < self.min_samples_leaf <= 0.5:
-
  /*else*/ {
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 276, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = PyObject_RichCompare(__pyx_float_0_, __pyx_t_2, Py_LT); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 276, __pyx_L1_error)
-    if (__Pyx_PyObject_IsTrue(__pyx_t_1)) {
-      __Pyx_DECREF(__pyx_t_1);
-      __pyx_t_1 = PyObject_RichCompare(__pyx_t_2, __pyx_float_0_5, Py_LE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 276, __pyx_L1_error)
-    }
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 276, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_9 = ((!__pyx_t_7) != 0);
-    if (unlikely(__pyx_t_9)) {
-/* … */
-    }
-
+277:                 raise ValueError("min_samples_leaf must be at least 1 "
-
      __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_builtin_ValueError, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 277, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_Raise(__pyx_t_1, 0, 0, 0);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __PYX_ERR(0, 277, __pyx_L1_error)
-
 278:                                  "or in (0, 0.5], got %s"
-
+279:                                  % self.min_samples_leaf)
-
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 279, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_2 = __Pyx_PyString_FormatSafe(__pyx_kp_s_min_samples_leaf_must_be_at_leas, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 279, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+280:             min_samples_leaf = int(ceil(self.min_samples_leaf * n_samples))
-
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_ceil); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 280, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 280, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_3 = PyNumber_Multiply(__pyx_t_6, __pyx_v_n_samples); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 280, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
-      if (likely(__pyx_t_6)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-        __Pyx_INCREF(__pyx_t_6);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_2, function);
-      }
-    }
-    __pyx_t_1 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_6, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
-    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 280, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyNumber_Int(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 280, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_v_min_samples_leaf = __pyx_t_2;
-    __pyx_t_2 = 0;
-  }
-  __pyx_L13:;
-
+281:         max_depth = (np.iinfo(np.int32).max if self.max_depth is None
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_9 = (__pyx_t_1 == Py_None);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if ((__pyx_t_9 != 0)) {
-    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_iinfo); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_int32); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
-      __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_6);
-      if (likely(__pyx_t_3)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-        __Pyx_INCREF(__pyx_t_3);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_6, function);
-      }
-    }
-    __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_3, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_4);
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_max); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_2 = __pyx_t_6;
-    __pyx_t_6 = 0;
-  } else {
-
+282:                      else self.max_depth)
-
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 282, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_2 = __pyx_t_6;
-    __pyx_t_6 = 0;
-  }
-  __pyx_v_max_depth = __pyx_t_2;
-  __pyx_t_2 = 0;
-
 283:
-
+284:         self.tree = Tree(
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_Tree); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 284, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-/* … */
-  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 284, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_tree, __pyx_t_12) < 0) __PYX_ERR(0, 284, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-
+285:             n_features = n_features,
-
  __pyx_t_6 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 285, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_features, __pyx_v_n_features) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
-
 286:             # line below is taken from DecisionTreeRegressor.fit method source
-
 287:             #   which comments that the tree shouldn't need the n_classes parameter
-
 288:             #   but it apparently does
-
+289:             n_classes = np.array([1] * n_outputs, dtype=np.intp),
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_array); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_int_1);
-  __Pyx_GIVEREF(__pyx_int_1);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_int_1);
-  { PyObject* __pyx_temp = PyNumber_InPlaceMultiply(__pyx_t_1, __pyx_v_n_outputs); if (unlikely(!__pyx_temp)) __PYX_ERR(0, 289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_temp);
-    __Pyx_DECREF(__pyx_t_1);
-    __pyx_t_1 = __pyx_temp;
-  }
-  __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_GIVEREF(__pyx_t_1);
-  PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_1);
-  __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GetModuleGlobalName(__pyx_t_11, __pyx_n_s_np); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_intp); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_dtype, __pyx_t_12) < 0) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 289, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_classes, __pyx_t_12) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-
+290:             n_outputs = n_outputs)
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_n_outputs, __pyx_v_n_outputs) < 0) __PYX_ERR(0, 285, __pyx_L1_error)
-
+291:         splitter = BestSplitter(criterion = CausalMSE(1, X.shape[0]),
-
  __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_BestSplitter); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __pyx_t_6 = __Pyx_PyDict_NewPresized(5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_int_1);
-  __Pyx_GIVEREF(__pyx_int_1);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_int_1);
-  __Pyx_GIVEREF(__pyx_t_1);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_1);
-  __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_Call(((PyObject *)__pyx_ptype_8causalml_9inference_4tree_10causaltree_CausalMSE), __pyx_t_2, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_criterion, __pyx_t_1) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-/* … */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_v_splitter = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+292:             max_features = n_features,
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_max_features, __pyx_v_n_features) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
-
+293:             min_samples_leaf = min_samples_leaf,
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_weight_leaf, __pyx_int_0) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
-
 294:             min_weight_leaf = 0, # from DecisionTreeRegressor default
-
+295:             random_state = random_state)
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 291, __pyx_L1_error)
-
 296:         # hardcoded values below come from defaults values in
-
 297:         #   sklearn.tree._classes.DecisionTreeRegressor
-
+298:         builder = DepthFirstTreeBuilder(
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_DepthFirstTreeBuilder); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 298, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_6); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 298, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __pyx_v_builder = __pyx_t_12;
-  __pyx_t_12 = 0;
-
+299:             splitter = splitter,
-
  __pyx_t_6 = __Pyx_PyDict_NewPresized(7); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 299, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_splitter, __pyx_v_splitter) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_split, __pyx_int_2) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-
 300:             min_samples_split = 2,
-
+301:             min_samples_leaf = min_samples_leaf,
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_weight_leaf, __pyx_int_0) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-
 302:             min_weight_leaf = 0,
-
+303:             max_depth = max_depth,
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_impurity_decrease, __pyx_int_0) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-
 304:             min_impurity_decrease = 0,
-
+305:             min_impurity_split = float("-inf"))
-
  __pyx_t_12 = __Pyx_PyNumber_Float(__pyx_kp_s_inf); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 305, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_min_impurity_split, __pyx_t_12) < 0) __PYX_ERR(0, 299, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-
+306:         builder.build(
-
  __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_v_builder, __pyx_n_s_build); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-/* … */
-  __pyx_t_1 = PyTuple_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GIVEREF(__pyx_t_6);
-  PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_6);
-  __pyx_t_6 = 0;
-/* … */
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_t_1, __pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+307:             self.tree,
-
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 307, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-
+308:             X = X,
-
  __pyx_t_6 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_X, __pyx_v_X) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
-
+309:             y = y,
-
  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_y, __pyx_v_y) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
-
+310:             sample_weight = 1 + self.eps * w)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_eps); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 310, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = PyNumber_Multiply(__pyx_t_2, __pyx_v_w); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 310, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyInt_AddCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 310, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (PyDict_SetItem(__pyx_t_6, __pyx_n_s_sample_weight, __pyx_t_2) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 311:
-
+312:         return self
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_self);
-  __pyx_r = __pyx_v_self;
-  goto __pyx_L0;
-
 313:
-
+314:     def predict(self, X):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict[] = "Predict treatment effects.\n\n        Args:\n            X (np.matrix): a feature matrix\n\n        Returns:\n            (numpy.ndarray): Predictions of treatment effects.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,0};
-    PyObject* values[2] = {0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, 1); __PYX_ERR(0, 314, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 314, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 314, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict(__pyx_self, __pyx_v_self, __pyx_v_X);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_4predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X) {
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict", 0);
-  __Pyx_INCREF(__pyx_v_X);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__31 = PyTuple_Pack(2, __pyx_n_s_self, __pyx_n_s_X); if (unlikely(!__pyx_tuple__31)) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__31);
-  __Pyx_GIVEREF(__pyx_tuple__31);
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_5predict, 0, __pyx_n_s_CausalTreeRegressor_predict, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__32)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_predict, __pyx_t_1) < 0) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_codeobj__32 = (PyObject*)__Pyx_PyCode_New(2, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__31, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_predict, 314, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__32)) __PYX_ERR(0, 314, __pyx_L1_error)
-
 315:         """Predict treatment effects.
-
 316:
-
 317:         Args:
-
 318:             X (np.matrix): a feature matrix
-
 319:
-
 320:         Returns:
-
 321:             (numpy.ndarray): Predictions of treatment effects.
-
 322:         """
-
+323:         X = check_array(X, dtype=DTYPE, accept_sparse="csr")
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_check_array); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_v_X);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
-  __pyx_t_3 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_DTYPE); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_dtype, __pyx_t_4) < 0) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_accept_sparse, __pyx_n_s_csr) < 0) __PYX_ERR(0, 323, __pyx_L1_error)
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_4);
-  __pyx_t_4 = 0;
-
+324:         return self.tree.predict(X).reshape((-1, 1))
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_predict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_3 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_2, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_X);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_reshape); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_tuple__4) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_tuple__4);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_r = __pyx_t_4;
-  __pyx_t_4 = 0;
-  goto __pyx_L0;
-
 325:
-
+326:     def fit_predict(self, X, treatment, y, return_ci=False, n_bootstraps=1000, bootstrap_size=10000, verbose=False):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict[] = "Fit the Causal Tree model and predict treatment effects.\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n            return_ci (bool): whether to return confidence intervals\n            n_bootstraps (int): number of bootstrap iterations\n            bootstrap_size (int): number of samples per bootstrap\n            verbose (str): whether to output progress logs\n\n        Returns:\n           (tuple):\n\n             - te (numpy.ndarray): Predictions of treatment effects.\n             - te_lower (numpy.ndarray, optional): lower bounds of treatment effects\n             - te_upper (numpy.ndarray, optional): upper bounds of treatment effects\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict = {"fit_predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_return_ci = 0;
-  PyObject *__pyx_v_n_bootstraps = 0;
-  PyObject *__pyx_v_bootstrap_size = 0;
-  PyObject *__pyx_v_verbose = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit_predict (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_return_ci,&__pyx_n_s_n_bootstraps,&__pyx_n_s_bootstrap_size,&__pyx_n_s_verbose,0};
-    PyObject* values[8] = {0,0,0,0,0,0,0,0};
-    values[4] = ((PyObject *)((PyObject *)Py_False));
-    values[5] = ((PyObject *)((PyObject *)__pyx_int_1000));
-    values[6] = ((PyObject *)((PyObject *)__pyx_int_10000));
-    values[7] = ((PyObject *)((PyObject *)Py_False));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 1); __PYX_ERR(0, 326, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 2); __PYX_ERR(0, 326, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, 3); __PYX_ERR(0, 326, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_return_ci);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_bootstraps);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_bootstrap_size);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_verbose);
-          if (value) { values[7] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit_predict") < 0)) __PYX_ERR(0, 326, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_return_ci = values[4];
-    __pyx_v_n_bootstraps = values[5];
-    __pyx_v_bootstrap_size = values[6];
-    __pyx_v_verbose = values[7];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fit_predict", 0, 4, 8, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 326, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit_predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_return_ci, __pyx_v_n_bootstraps, __pyx_v_bootstrap_size, __pyx_v_verbose);
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_6fit_predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_return_ci, PyObject *__pyx_v_n_bootstraps, PyObject *__pyx_v_bootstrap_size, PyObject *__pyx_v_verbose) {
-  PyObject *__pyx_v_te = NULL;
-  PyObject *__pyx_v_start = NULL;
-  PyObject *__pyx_v_te_bootstraps = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_te_b = NULL;
-  PyObject *__pyx_v_now = NULL;
-  PyObject *__pyx_v_lapsed = NULL;
-  PyObject *__pyx_v_te_lower = NULL;
-  PyObject *__pyx_v_te_upper = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit_predict", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_XDECREF(__pyx_t_12);
-  __Pyx_XDECREF(__pyx_t_13);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.fit_predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_te);
-  __Pyx_XDECREF(__pyx_v_start);
-  __Pyx_XDECREF(__pyx_v_te_bootstraps);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_te_b);
-  __Pyx_XDECREF(__pyx_v_now);
-  __Pyx_XDECREF(__pyx_v_lapsed);
-  __Pyx_XDECREF(__pyx_v_te_lower);
-  __Pyx_XDECREF(__pyx_v_te_upper);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__33 = PyTuple_Pack(17, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_return_ci, __pyx_n_s_n_bootstraps, __pyx_n_s_bootstrap_size, __pyx_n_s_verbose, __pyx_n_s_te, __pyx_n_s_start, __pyx_n_s_te_bootstraps, __pyx_n_s_i, __pyx_n_s_te_b, __pyx_n_s_now, __pyx_n_s_lapsed, __pyx_n_s_te_lower, __pyx_n_s_te_upper); if (unlikely(!__pyx_tuple__33)) __PYX_ERR(0, 326, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__33);
-  __Pyx_GIVEREF(__pyx_tuple__33);
-  __pyx_codeobj__34 = (PyObject*)__Pyx_PyCode_New(8, 0, 17, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__33, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_fit_predict, 326, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__34)) __PYX_ERR(0, 326, __pyx_L1_error)
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_7fit_predict, 0, __pyx_n_s_CausalTreeRegressor_fit_predict, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__34)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 326, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__35);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_fit_predict, __pyx_t_1) < 0) __PYX_ERR(0, 326, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_tuple__35 = PyTuple_Pack(4, ((PyObject *)Py_False), ((PyObject *)__pyx_int_1000), ((PyObject *)__pyx_int_10000), ((PyObject *)Py_False)); if (unlikely(!__pyx_tuple__35)) __PYX_ERR(0, 326, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__35);
-  __Pyx_GIVEREF(__pyx_tuple__35);
-
 327:         """Fit the Causal Tree model and predict treatment effects.
-
 328:
-
 329:         Args:
-
 330:             X (np.matrix): a feature matrix
-
 331:             treatment (np.array): a treatment vector
-
 332:             y (np.array): an outcome vector
-
 333:             return_ci (bool): whether to return confidence intervals
-
 334:             n_bootstraps (int): number of bootstrap iterations
-
 335:             bootstrap_size (int): number of samples per bootstrap
-
 336:             verbose (str): whether to output progress logs
-
 337:
-
 338:         Returns:
-
 339:            (tuple):
-
 340:
-
 341:              - te (numpy.ndarray): Predictions of treatment effects.
-
 342:              - te_lower (numpy.ndarray, optional): lower bounds of treatment effects
-
 343:              - te_upper (numpy.ndarray, optional): upper bounds of treatment effects
-
 344:         """
-
+345:         self.fit(X, treatment, y)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 345, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 345, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_X);
-    __Pyx_GIVEREF(__pyx_v_X);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
-    __Pyx_INCREF(__pyx_v_treatment);
-    __Pyx_GIVEREF(__pyx_v_treatment);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_treatment);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 345, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+346:         te = self.predict(X)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 346, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 346, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_te = __pyx_t_1;
-  __pyx_t_1 = 0;
-
 347:
-
+348:         if not return_ci:
-
  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_v_return_ci); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 348, __pyx_L1_error)
-  __pyx_t_7 = ((!__pyx_t_6) != 0);
-  if (__pyx_t_7) {
-/* … */
-  }
-
+349:             return te
-
    __Pyx_XDECREF(__pyx_r);
-    __Pyx_INCREF(__pyx_v_te);
-    __pyx_r = __pyx_v_te;
-    goto __pyx_L0;
-
 350:         else:
-
+351:             start = pd.datetime.today()
-
  /*else*/ {
-    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_pd); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 351, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_datetime); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 351, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_today); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 351, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = NULL;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-      if (likely(__pyx_t_5)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-        __Pyx_INCREF(__pyx_t_5);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_2, function);
-      }
-    }
-    __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 351, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_v_start = __pyx_t_1;
-    __pyx_t_1 = 0;
-
+352:             te_bootstraps = np.zeros(shape=(X.shape[0], n_bootstraps))
-
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_zeros); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_5, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_GIVEREF(__pyx_t_3);
-    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3);
-    __Pyx_INCREF(__pyx_v_n_bootstraps);
-    __Pyx_GIVEREF(__pyx_v_n_bootstraps);
-    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_n_bootstraps);
-    __pyx_t_3 = 0;
-    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_shape, __pyx_t_5) < 0) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 352, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_v_te_bootstraps = __pyx_t_5;
-    __pyx_t_5 = 0;
-
+353:             for i in range(n_bootstraps):
-
    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_v_n_bootstraps); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (likely(PyList_CheckExact(__pyx_t_5)) || PyTuple_CheckExact(__pyx_t_5)) {
-      __pyx_t_1 = __pyx_t_5; __Pyx_INCREF(__pyx_t_1); __pyx_t_8 = 0;
-      __pyx_t_9 = NULL;
-    } else {
-      __pyx_t_8 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 353, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_9 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 353, __pyx_L1_error)
-    }
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    for (;;) {
-      if (likely(!__pyx_t_9)) {
-        if (likely(PyList_CheckExact(__pyx_t_1))) {
-          if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_1)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_8); __Pyx_INCREF(__pyx_t_5); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 353, __pyx_L1_error)
-          #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_1, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_5);
-          #endif
-        } else {
-          if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_8); __Pyx_INCREF(__pyx_t_5); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 353, __pyx_L1_error)
-          #else
-          __pyx_t_5 = PySequence_ITEM(__pyx_t_1, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 353, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_5);
-          #endif
-        }
-      } else {
-        __pyx_t_5 = __pyx_t_9(__pyx_t_1);
-        if (unlikely(!__pyx_t_5)) {
-          PyObject* exc_type = PyErr_Occurred();
-          if (exc_type) {
-            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 353, __pyx_L1_error)
-          }
-          break;
-        }
-        __Pyx_GOTREF(__pyx_t_5);
-      }
-      __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_5);
-      __pyx_t_5 = 0;
-/* … */
-    }
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+354:                 te_b = self.bootstrap(X, treatment, y, size=bootstrap_size)
-
      __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_bootstrap); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 354, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 354, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_INCREF(__pyx_v_X);
-      __Pyx_GIVEREF(__pyx_v_X);
-      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
-      __Pyx_INCREF(__pyx_v_treatment);
-      __Pyx_GIVEREF(__pyx_v_treatment);
-      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_treatment);
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_y);
-      __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 354, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_size, __pyx_v_bootstrap_size) < 0) __PYX_ERR(0, 354, __pyx_L1_error)
-      __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 354, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_te_b, __pyx_t_10);
-      __pyx_t_10 = 0;
-
+355:                 te_bootstraps[:,i] = np.ravel(te_b)
-
      __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 355, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_ravel); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 355, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_3 = NULL;
-      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-        __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-        if (likely(__pyx_t_3)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-          __Pyx_INCREF(__pyx_t_3);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_2, function);
-        }
-      }
-      __pyx_t_10 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_te_b) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_te_b);
-      __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-      if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 355, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 355, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_INCREF(__pyx_slice__5);
-      __Pyx_GIVEREF(__pyx_slice__5);
-      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_slice__5);
-      __Pyx_INCREF(__pyx_v_i);
-      __Pyx_GIVEREF(__pyx_v_i);
-      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_i);
-      if (unlikely(PyObject_SetItem(__pyx_v_te_bootstraps, __pyx_t_2, __pyx_t_10) < 0)) __PYX_ERR(0, 355, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-/* … */
-  __pyx_slice__5 = PySlice_New(Py_None, Py_None, Py_None); if (unlikely(!__pyx_slice__5)) __PYX_ERR(0, 355, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_slice__5);
-  __Pyx_GIVEREF(__pyx_slice__5);
-
+356:                 if verbose:
-
      __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_v_verbose); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 356, __pyx_L1_error)
-      if (__pyx_t_7) {
-/* … */
-      }
-
+357:                     now = pd.datetime.today()
-
        __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_pd); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 357, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_datetime); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 357, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_today); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 357, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-          __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-          if (likely(__pyx_t_3)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-            __Pyx_INCREF(__pyx_t_3);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_2, function);
-          }
-        }
-        __pyx_t_10 = (__pyx_t_3) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-        if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 357, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_now, __pyx_t_10);
-        __pyx_t_10 = 0;
-
+358:                     lapsed = (now-start).seconds / 60
-
        __pyx_t_10 = PyNumber_Subtract(__pyx_v_now, __pyx_v_start); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 358, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_10, __pyx_n_s_seconds); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 358, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-        __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_int_60); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 358, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_lapsed, __pyx_t_10);
-        __pyx_t_10 = 0;
-
+359:                     logger.info('{}/{} bootstraps completed. ({:.01f} min lapsed)'.format(i+1, n_bootstraps, lapsed))
-
        __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_logger); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_info); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 359, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_kp_s_bootstraps_completed_01f_min_la, __pyx_n_s_format); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 359, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __pyx_t_11 = __Pyx_PyInt_AddObjC(__pyx_v_i, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 359, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_12 = NULL;
-        __pyx_t_4 = 0;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
-          __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_5);
-          if (likely(__pyx_t_12)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-            __Pyx_INCREF(__pyx_t_12);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_5, function);
-            __pyx_t_4 = 1;
-          }
-        }
-        #if CYTHON_FAST_PYCALL
-        if (PyFunction_Check(__pyx_t_5)) {
-          PyObject *__pyx_temp[4] = {__pyx_t_12, __pyx_t_11, __pyx_v_n_bootstraps, __pyx_v_lapsed};
-          __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-          __Pyx_GOTREF(__pyx_t_2);
-          __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        } else
-        #endif
-        #if CYTHON_FAST_PYCCALL
-        if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
-          PyObject *__pyx_temp[4] = {__pyx_t_12, __pyx_t_11, __pyx_v_n_bootstraps, __pyx_v_lapsed};
-          __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-          __Pyx_GOTREF(__pyx_t_2);
-          __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        } else
-        #endif
-        {
-          __pyx_t_13 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 359, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_13);
-          if (__pyx_t_12) {
-            __Pyx_GIVEREF(__pyx_t_12); PyTuple_SET_ITEM(__pyx_t_13, 0, __pyx_t_12); __pyx_t_12 = NULL;
-          }
-          __Pyx_GIVEREF(__pyx_t_11);
-          PyTuple_SET_ITEM(__pyx_t_13, 0+__pyx_t_4, __pyx_t_11);
-          __Pyx_INCREF(__pyx_v_n_bootstraps);
-          __Pyx_GIVEREF(__pyx_v_n_bootstraps);
-          PyTuple_SET_ITEM(__pyx_t_13, 1+__pyx_t_4, __pyx_v_n_bootstraps);
-          __Pyx_INCREF(__pyx_v_lapsed);
-          __Pyx_GIVEREF(__pyx_v_lapsed);
-          PyTuple_SET_ITEM(__pyx_t_13, 2+__pyx_t_4, __pyx_v_lapsed);
-          __pyx_t_11 = 0;
-          __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_13, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 359, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_2);
-          __Pyx_DECREF(__pyx_t_13); __pyx_t_13 = 0;
-        }
-        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        __pyx_t_5 = NULL;
-        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-          __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_3);
-          if (likely(__pyx_t_5)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-            __Pyx_INCREF(__pyx_t_5);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_3, function);
-          }
-        }
-        __pyx_t_10 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_5, __pyx_t_2) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_2);
-        __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 359, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
 360:
-
+361:             te_lower = np.percentile(te_bootstraps, (self.ate_alpha/2)*100, axis=1)
-
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_percentile); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = PyNumber_Multiply(__pyx_t_3, __pyx_int_100); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = PyTuple_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_INCREF(__pyx_v_te_bootstraps);
-    __Pyx_GIVEREF(__pyx_v_te_bootstraps);
-    PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_te_bootstraps);
-    __Pyx_GIVEREF(__pyx_t_1);
-    PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_t_1);
-    __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 361, __pyx_L1_error)
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 361, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_v_te_lower = __pyx_t_2;
-    __pyx_t_2 = 0;
-
+362:             te_upper = np.percentile(te_bootstraps, (1 - self.ate_alpha/2)*100, axis=1)
-
    __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_percentile); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = PyNumber_Multiply(__pyx_t_2, __pyx_int_100); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_v_te_bootstraps);
-    __Pyx_GIVEREF(__pyx_v_te_bootstraps);
-    PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_te_bootstraps);
-    __Pyx_GIVEREF(__pyx_t_3);
-    PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 362, __pyx_L1_error)
-    __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 362, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_v_te_upper = __pyx_t_10;
-    __pyx_t_10 = 0;
-
 363:
-
+364:             return (te, te_lower, te_upper)
-
    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_10 = PyTuple_New(3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 364, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __Pyx_INCREF(__pyx_v_te);
-    __Pyx_GIVEREF(__pyx_v_te);
-    PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_v_te);
-    __Pyx_INCREF(__pyx_v_te_lower);
-    __Pyx_GIVEREF(__pyx_v_te_lower);
-    PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_v_te_lower);
-    __Pyx_INCREF(__pyx_v_te_upper);
-    __Pyx_GIVEREF(__pyx_v_te_upper);
-    PyTuple_SET_ITEM(__pyx_t_10, 2, __pyx_v_te_upper);
-    __pyx_r = __pyx_t_10;
-    __pyx_t_10 = 0;
-    goto __pyx_L0;
-  }
-
 365:
-
+366:     def bootstrap(self, X, treatment, y, size=10000):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap[] = "Runs a single bootstrap.\n\n        Fits on bootstrapped sample, then predicts on whole population.\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n            size (int, optional): bootstrap sample size\n\n        Returns:\n            (np.array): bootstrap predictions\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap = {"bootstrap", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_size = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("bootstrap (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_size,0};
-    PyObject* values[5] = {0,0,0,0,0};
-    values[4] = ((PyObject *)((PyObject *)__pyx_int_10000));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 1); __PYX_ERR(0, 366, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 2); __PYX_ERR(0, 366, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, 3); __PYX_ERR(0, 366, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_size);
-          if (value) { values[4] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "bootstrap") < 0)) __PYX_ERR(0, 366, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_size = values[4];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("bootstrap", 0, 4, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 366, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_size);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_8bootstrap(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_size) {
-  PyObject *__pyx_v_idxs = NULL;
-  PyObject *__pyx_v_X_b = NULL;
-  PyObject *__pyx_v_treatment_b = NULL;
-  PyObject *__pyx_v_y_b = NULL;
-  PyObject *__pyx_v_te_b = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("bootstrap", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_idxs);
-  __Pyx_XDECREF(__pyx_v_X_b);
-  __Pyx_XDECREF(__pyx_v_treatment_b);
-  __Pyx_XDECREF(__pyx_v_y_b);
-  __Pyx_XDECREF(__pyx_v_te_b);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__36 = PyTuple_Pack(10, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_size, __pyx_n_s_idxs, __pyx_n_s_X_b, __pyx_n_s_treatment_b, __pyx_n_s_y_b, __pyx_n_s_te_b); if (unlikely(!__pyx_tuple__36)) __PYX_ERR(0, 366, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__36);
-  __Pyx_GIVEREF(__pyx_tuple__36);
-  __pyx_codeobj__37 = (PyObject*)__Pyx_PyCode_New(5, 0, 10, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__36, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_bootstrap, 366, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__37)) __PYX_ERR(0, 366, __pyx_L1_error)
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_9bootstrap, 0, __pyx_n_s_CausalTreeRegressor_bootstrap, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__37)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 366, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_1, __pyx_tuple__38);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_bootstrap, __pyx_t_1) < 0) __PYX_ERR(0, 366, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_tuple__38 = PyTuple_Pack(1, ((PyObject *)__pyx_int_10000)); if (unlikely(!__pyx_tuple__38)) __PYX_ERR(0, 366, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__38);
-  __Pyx_GIVEREF(__pyx_tuple__38);
-
 367:         """Runs a single bootstrap.
-
 368:
-
 369:         Fits on bootstrapped sample, then predicts on whole population.
-
 370:
-
 371:         Args:
-
 372:             X (np.matrix): a feature matrix
-
 373:             treatment (np.array): a treatment vector
-
 374:             y (np.array): an outcome vector
-
 375:             size (int, optional): bootstrap sample size
-
 376:
-
 377:         Returns:
-
 378:             (np.array): bootstrap predictions
-
 379:         """
-
+380:         idxs = np.random.choice(np.arange(0, X.shape[0]), size=size)
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_np); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_random); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_choice); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_arange); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_3, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  __pyx_t_6 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_4);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_4, function);
-      __pyx_t_6 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_4)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_int_0, __pyx_t_5};
-    __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_4)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_int_0, __pyx_t_5};
-    __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_4, __pyx_temp+1-__pyx_t_6, 2+__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  } else
-  #endif
-  {
-    __pyx_t_7 = PyTuple_New(2+__pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 380, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_int_0);
-    __Pyx_GIVEREF(__pyx_int_0);
-    PyTuple_SET_ITEM(__pyx_t_7, 0+__pyx_t_6, __pyx_int_0);
-    __Pyx_GIVEREF(__pyx_t_5);
-    PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_6, __pyx_t_5);
-    __pyx_t_5 = 0;
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_7, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = PyTuple_New(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_GIVEREF(__pyx_t_2);
-  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_2);
-  __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_size, __pyx_v_size) < 0) __PYX_ERR(0, 380, __pyx_L1_error)
-  __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 380, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_idxs = __pyx_t_7;
-  __pyx_t_7 = 0;
-
+381:         X_b = X[idxs]
-
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 381, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_v_X_b = __pyx_t_7;
-  __pyx_t_7 = 0;
-
+382:         treatment_b = treatment[idxs]
-
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 382, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_v_treatment_b = __pyx_t_7;
-  __pyx_t_7 = 0;
-
+383:         y_b = y[idxs]
-
  __pyx_t_7 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_idxs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 383, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_v_y_b = __pyx_t_7;
-  __pyx_t_7 = 0;
-
+384:         self.fit(X=X_b, treatment=treatment_b, y=y_b)
-
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 384, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 384, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_X_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_treatment, __pyx_v_treatment_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_y, __pyx_v_y_b) < 0) __PYX_ERR(0, 384, __pyx_L1_error)
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 384, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+385:         te_b = self.predict(X=X)
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_predict); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_X) < 0) __PYX_ERR(0, 385, __pyx_L1_error)
-  __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_te_b = __pyx_t_7;
-  __pyx_t_7 = 0;
-
+386:         return te_b
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_te_b);
-  __pyx_r = __pyx_v_te_b;
-  goto __pyx_L0;
-
 387:
-
+388:     def estimate_ate(self, X, treatment, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate[] = "Estimate the Average Treatment Effect (ATE).\n\n        Args:\n            X (np.matrix): a feature matrix\n            treatment (np.array): a treatment vector\n            y (np.array): an outcome vector\n\n        Returns:\n            The mean and confidence interval (LB, UB) of the ATE estimate.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate = {"estimate_ate", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("estimate_ate (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 1); __PYX_ERR(0, 388, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 2); __PYX_ERR(0, 388, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, 3); __PYX_ERR(0, 388, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "estimate_ate") < 0)) __PYX_ERR(0, 388, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("estimate_ate", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 388, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.estimate_ate", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_10estimate_ate(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_dhat = NULL;
-  PyObject *__pyx_v_te = NULL;
-  PyObject *__pyx_v_se = NULL;
-  PyObject *__pyx_v_te_lb = NULL;
-  PyObject *__pyx_v_te_ub = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("estimate_ate", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.causaltree.CausalTreeRegressor.estimate_ate", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_dhat);
-  __Pyx_XDECREF(__pyx_v_te);
-  __Pyx_XDECREF(__pyx_v_se);
-  __Pyx_XDECREF(__pyx_v_te_lb);
-  __Pyx_XDECREF(__pyx_v_te_ub);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__39 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_dhat, __pyx_n_s_te, __pyx_n_s_se, __pyx_n_s_te_lb, __pyx_n_s_te_ub); if (unlikely(!__pyx_tuple__39)) __PYX_ERR(0, 388, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__39);
-  __Pyx_GIVEREF(__pyx_tuple__39);
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_10causaltree_19CausalTreeRegressor_11estimate_ate, 0, __pyx_n_s_CausalTreeRegressor_estimate_ate, NULL, __pyx_n_s_causalml_inference_tree_causaltr, __pyx_d, ((PyObject *)__pyx_codeobj__40)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 388, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (__Pyx_SetNameInClass(__pyx_t_2, __pyx_n_s_estimate_ate, __pyx_t_1) < 0) __PYX_ERR(0, 388, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_codeobj__40 = (PyObject*)__Pyx_PyCode_New(4, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__39, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_causaltr_2, __pyx_n_s_estimate_ate, 388, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__40)) __PYX_ERR(0, 388, __pyx_L1_error)
-
 389:         """Estimate the Average Treatment Effect (ATE).
-
 390:
-
 391:         Args:
-
 392:             X (np.matrix): a feature matrix
-
 393:             treatment (np.array): a treatment vector
-
 394:             y (np.array): an outcome vector
-
 395:
-
 396:         Returns:
-
 397:             The mean and confidence interval (LB, UB) of the ATE estimate.
-
 398:         """
-
+399:         dhat = self.fit_predict(X, treatment, y)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fit_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 399, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[4] = {__pyx_t_3, __pyx_v_X, __pyx_v_treatment, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 3+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(3+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 399, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_X);
-    __Pyx_GIVEREF(__pyx_v_X);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
-    __Pyx_INCREF(__pyx_v_treatment);
-    __Pyx_GIVEREF(__pyx_v_treatment);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_treatment);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 399, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_dhat = __pyx_t_1;
-  __pyx_t_1 = 0;
-
 400:
-
+401:         te = dhat.mean()
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_dhat, __pyx_n_s_mean); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 401, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 401, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_te = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+402:         se = dhat.std() / X.shape[0]
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_dhat, __pyx_n_s_std); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_se = __pyx_t_2;
-  __pyx_t_2 = 0;
-
 403:
-
+404:         te_lb = te - se * norm.ppf(1 - self.ate_alpha / 2)
-
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_norm); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_ppf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_t_5) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = PyNumber_Multiply(__pyx_v_se, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyNumber_Subtract(__pyx_v_te, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 404, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_te_lb = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+405:         te_ub = te + se * norm.ppf(1 - self.ate_alpha / 2)
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_norm); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_ppf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_ate_alpha); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_int_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_t_3, 1, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_5);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_5, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_3, __pyx_t_1) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = PyNumber_Multiply(__pyx_v_se, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyNumber_Add(__pyx_v_te, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 405, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_te_ub = __pyx_t_2;
-  __pyx_t_2 = 0;
-
 406:
-
+407:         return te, te_lb, te_ub
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 407, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_te);
-  __Pyx_GIVEREF(__pyx_v_te);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_te);
-  __Pyx_INCREF(__pyx_v_te_lb);
-  __Pyx_GIVEREF(__pyx_v_te_lb);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_te_lb);
-  __Pyx_INCREF(__pyx_v_te_ub);
-  __Pyx_GIVEREF(__pyx_v_te_ub);
-  PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_te_ub);
-  __pyx_r = __pyx_t_2;
-  __pyx_t_2 = 0;
-  goto __pyx_L0;
-
diff --git a/causalml/inference/tree/uplift.html b/causalml/inference/tree/uplift.html deleted file mode 100644 index eb93d071..00000000 --- a/causalml/inference/tree/uplift.html +++ /dev/null @@ -1,13953 +0,0 @@ - - - - - - Cython: uplift.pyx - - - -

Generated by Cython 0.29.24

-

- Yellow lines hint at Python interaction.
- Click on a line that starts with a "+" to see the C code that Cython generated for it. -

-

Raw output: uplift.c

-
+0001: # cython: cdivision=True
-
  __pyx_t_1 = __Pyx_PyDict_NewPresized(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_test, __pyx_t_1) < 0) __PYX_ERR(0, 1, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0002: # cython: boundscheck=False
-
 0003: # cython: wraparound=False
-
 0004: """
-
 0005: Forest of trees-based ensemble methods for Uplift modeling on Classification
-
 0006: Problem. Those methods include random forests and extremely randomized trees.
-
 0007:
-
 0008: The module structure is the following:
-
 0009: - The ``UpliftRandomForestClassifier`` base class implements different
-
 0010:   variants of uplift models based on random forest, with 'fit' and 'predict'
-
 0011:   method.
-
 0012: - The ``UpliftTreeClassifier`` base class implements the uplift trees (without
-
 0013:   Bootstrapping for random forest), this class is called within
-
 0014:   ``UpliftRandomForestClassifier`` for constructing random forest.
-
 0015: """
-
 0016:
-
 0017: # Authors: Zhenyu Zhao <zhenyuz@uber.com>
-
 0018: #          Totte Harinen <totte@uber.com>
-
 0019:
-
+0020: from collections import defaultdict
-
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_defaultdict);
-  __Pyx_GIVEREF(__pyx_n_s_defaultdict);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_defaultdict);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_collections, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_defaultdict, __pyx_t_1) < 0) __PYX_ERR(0, 20, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0021: import cython
-
+0022: from joblib import Parallel, delayed
-
  __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s_Parallel);
-  __Pyx_GIVEREF(__pyx_n_s_Parallel);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_Parallel);
-  __Pyx_INCREF(__pyx_n_s_delayed);
-  __Pyx_GIVEREF(__pyx_n_s_delayed);
-  PyList_SET_ITEM(__pyx_t_2, 1, __pyx_n_s_delayed);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_joblib, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_Parallel, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_delayed); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_delayed, __pyx_t_2) < 0) __PYX_ERR(0, 22, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0023: import multiprocessing as mp
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_multiprocessing, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 23, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_mp, __pyx_t_1) < 0) __PYX_ERR(0, 23, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0024: cimport numpy as np
-
+0025: import numpy as np
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_numpy, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 25, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_np, __pyx_t_1) < 0) __PYX_ERR(0, 25, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0026: from packaging import version
-
  __pyx_t_1 = PyList_New(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 26, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_version);
-  __Pyx_GIVEREF(__pyx_n_s_version);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_version);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_packaging, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 26, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_version); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 26, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_version, __pyx_t_1) < 0) __PYX_ERR(0, 26, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0027: import pandas as pd
-
  __pyx_t_2 = __Pyx_Import(__pyx_n_s_pandas, 0, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 27, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_pd, __pyx_t_2) < 0) __PYX_ERR(0, 27, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0028: import scipy.stats as stats
-
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 28, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s__14);
-  __Pyx_GIVEREF(__pyx_n_s__14);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s__14);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_scipy_stats, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 28, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_stats, __pyx_t_1) < 0) __PYX_ERR(0, 28, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0029: import sklearn
-
  __pyx_t_1 = __Pyx_Import(__pyx_n_s_sklearn, 0, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 29, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_sklearn, __pyx_t_1) < 0) __PYX_ERR(0, 29, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0030: from sklearn.utils import check_array, check_random_state, check_X_y
-
  __pyx_t_1 = PyList_New(3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_INCREF(__pyx_n_s_check_array);
-  __Pyx_GIVEREF(__pyx_n_s_check_array);
-  PyList_SET_ITEM(__pyx_t_1, 0, __pyx_n_s_check_array);
-  __Pyx_INCREF(__pyx_n_s_check_random_state);
-  __Pyx_GIVEREF(__pyx_n_s_check_random_state);
-  PyList_SET_ITEM(__pyx_t_1, 1, __pyx_n_s_check_random_state);
-  __Pyx_INCREF(__pyx_n_s_check_X_y);
-  __Pyx_GIVEREF(__pyx_n_s_check_X_y);
-  PyList_SET_ITEM(__pyx_t_1, 2, __pyx_n_s_check_X_y);
-  __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_utils, __pyx_t_1, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_array); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_array, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_random_state, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_check_X_y, __pyx_t_1) < 0) __PYX_ERR(0, 30, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0031: from typing import List
-
  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 31, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_n_s_List);
-  __Pyx_GIVEREF(__pyx_n_s_List);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_List);
-  __pyx_t_1 = __Pyx_Import(__pyx_n_s_typing, __pyx_t_2, -1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 31, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_1, __pyx_n_s_List); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 31, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_List, __pyx_t_2) < 0) __PYX_ERR(0, 31, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0032: if version.parse(sklearn.__version__) >= version.parse('0.22.0'):
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_version); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_parse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_sklearn); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_version_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_version); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_parse); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_tuple__15, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyObject_RichCompare(__pyx_t_1, __pyx_t_3, Py_GE); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_4 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_4 < 0)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__pyx_t_4) {
-/* … */
-    goto __pyx_L2;
-  }
-/* … */
-  __pyx_tuple__15 = PyTuple_Pack(1, __pyx_kp_s_0_22_0); if (unlikely(!__pyx_tuple__15)) __PYX_ERR(0, 32, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__15);
-  __Pyx_GIVEREF(__pyx_tuple__15);
-
+0033:     from sklearn.utils._testing import ignore_warnings
-
    __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 33, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_n_s_ignore_warnings);
-    __Pyx_GIVEREF(__pyx_n_s_ignore_warnings);
-    PyList_SET_ITEM(__pyx_t_2, 0, __pyx_n_s_ignore_warnings);
-    __pyx_t_3 = __Pyx_Import(__pyx_n_s_sklearn_utils__testing, __pyx_t_2, -1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 33, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_ImportFrom(__pyx_t_3, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 33, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (PyDict_SetItem(__pyx_d, __pyx_n_s_ignore_warnings, __pyx_t_2) < 0) __PYX_ERR(0, 33, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
 0034: else:
-
+0035:     from sklearn.utils.testing import ignore_warnings
-
  /*else*/ {
-    __pyx_t_3 = PyList_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 35, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_INCREF(__pyx_n_s_ignore_warnings);
-    __Pyx_GIVEREF(__pyx_n_s_ignore_warnings);
-    PyList_SET_ITEM(__pyx_t_3, 0, __pyx_n_s_ignore_warnings);
-    __pyx_t_2 = __Pyx_Import(__pyx_n_s_sklearn_utils_testing, __pyx_t_3, -1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 35, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_ImportFrom(__pyx_t_2, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 35, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    if (PyDict_SetItem(__pyx_d, __pyx_n_s_ignore_warnings, __pyx_t_3) < 0) __PYX_ERR(0, 35, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  }
-  __pyx_L2:;
-
 0036:
-
 0037:
-
+0038: MAX_INT = np.iinfo(np.int32).max
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_iinfo); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_int32); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_max); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_MAX_INT, __pyx_t_1) < 0) __PYX_ERR(0, 38, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0039:
-
 0040:
-
 0041: cdef extern from "math.h":
-
 0042:     double log(double x) nogil
-
 0043:
-
 0044:
-
 0045: @cython.cfunc
-
+0046: def kl_divergence(pk: cython.float, qk: cython.float) -> cython.float:
-
static float __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(float __pyx_v_pk, float __pyx_v_qk) {
-  float __pyx_v_eps;
-  float __pyx_v_S;
-  float __pyx_r;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("kl_divergence", 0);
-/* … */
-  /* function exit code */
-  __pyx_L0:;
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
 0047:     '''
-
 0048:     Calculate KL Divergence for binary classification.
-
 0049:
-
 0050:     sum(np.array(pk) * np.log(np.array(pk) / np.array(qk)))
-
 0051:
-
 0052:     Args
-
 0053:     ----
-
 0054:     pk : float
-
 0055:         The probability of 1 in one distribution.
-
 0056:     qk : float
-
 0057:         The probability of 1 in the other distribution.
-
 0058:
-
 0059:     Returns
-
 0060:     -------
-
 0061:     S : float
-
 0062:         The KL divergence.
-
 0063:     '''
-
 0064:
-
+0065:     eps: cython.float = 1e-6
-
  __pyx_v_eps = 1e-6;
-
 0066:     S: cython.float
-
 0067:
-
+0068:     if qk == 0.:
-
  __pyx_t_1 = ((__pyx_v_qk == 0.) != 0);
-  if (__pyx_t_1) {
-/* … */
-  }
-
+0069:         return 0.
-
    __pyx_r = 0.;
-    goto __pyx_L0;
-
 0070:
-
+0071:     qk = min(max(qk, eps), 1 - eps)
-
  __pyx_t_2 = (1.0 - __pyx_v_eps);
-  __pyx_t_3 = __pyx_v_eps;
-  __pyx_t_4 = __pyx_v_qk;
-  if (((__pyx_t_3 > __pyx_t_4) != 0)) {
-    __pyx_t_5 = __pyx_t_3;
-  } else {
-    __pyx_t_5 = __pyx_t_4;
-  }
-  __pyx_t_3 = __pyx_t_5;
-  if (((__pyx_t_2 < __pyx_t_3) != 0)) {
-    __pyx_t_5 = __pyx_t_2;
-  } else {
-    __pyx_t_5 = __pyx_t_3;
-  }
-  __pyx_v_qk = __pyx_t_5;
-
 0072:
-
+0073:     if pk == 0.:
-
  __pyx_t_1 = ((__pyx_v_pk == 0.) != 0);
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L4;
-  }
-
+0074:         S = -log(1 - qk)
-
    __pyx_v_S = (-log((1.0 - __pyx_v_qk)));
-
+0075:     elif pk == 1.:
-
  __pyx_t_1 = ((__pyx_v_pk == 1.) != 0);
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L4;
-  }
-
+0076:         S = -log(qk)
-
    __pyx_v_S = (-log(__pyx_v_qk));
-
 0077:     else:
-
+0078:         S = pk * log(pk / qk) + (1 - pk) * log((1 - pk) / (1 - qk))
-
  /*else*/ {
-    __pyx_v_S = ((__pyx_v_pk * log((__pyx_v_pk / __pyx_v_qk))) + ((1.0 - __pyx_v_pk) * log(((1.0 - __pyx_v_pk) / (1.0 - __pyx_v_qk)))));
-  }
-  __pyx_L4:;
-
 0079:
-
+0080:     return S
-
  __pyx_r = __pyx_v_S;
-  goto __pyx_L0;
-
 0081:
-
 0082:
-
 0083: @cython.cfunc
-
+0084: def entropyH(p: cython.float, q: cython.float=-1.) -> cython.float:
-
static float __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(float __pyx_v_p, struct __pyx_opt_args_8causalml_9inference_4tree_6uplift_entropyH *__pyx_optional_args) {
-  float __pyx_v_q = __pyx_k_;
-  float __pyx_r;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("entropyH", 0);
-  if (__pyx_optional_args) {
-    if (__pyx_optional_args->__pyx_n > 0) {
-      __pyx_v_q = __pyx_optional_args->q;
-    }
-  }
-/* … */
-  /* function exit code */
-  __pyx_L0:;
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_k_ = (-1.);
-/* … */
-struct __pyx_opt_args_8causalml_9inference_4tree_6uplift_entropyH {
-  int __pyx_n;
-  float q;
-};
-
 0085:     '''
-
 0086:     Entropy
-
 0087:
-
 0088:     Entropy calculation for normalization.
-
 0089:
-
 0090:     Args
-
 0091:     ----
-
 0092:     p : float
-
 0093:         The probability used in the entropy calculation.
-
 0094:
-
 0095:     q : float, optional, (default = -1.)
-
 0096:         The second probability used in the entropy calculation.
-
 0097:
-
 0098:     Returns
-
 0099:     -------
-
 0100:     entropy : float
-
 0101:     '''
-
 0102:
-
+0103:     if q == -1. and p > 0.:
-
  __pyx_t_2 = ((__pyx_v_q == -1.) != 0);
-  if (__pyx_t_2) {
-  } else {
-    __pyx_t_1 = __pyx_t_2;
-    goto __pyx_L4_bool_binop_done;
-  }
-  __pyx_t_2 = ((__pyx_v_p > 0.) != 0);
-  __pyx_t_1 = __pyx_t_2;
-  __pyx_L4_bool_binop_done:;
-  if (__pyx_t_1) {
-/* … */
-  }
-
+0104:         return -p * log(p)
-
    __pyx_r = ((-__pyx_v_p) * log(__pyx_v_p));
-    goto __pyx_L0;
-
+0105:     elif q > 0.:
-
  __pyx_t_1 = ((__pyx_v_q > 0.) != 0);
-  if (__pyx_t_1) {
-/* … */
-  }
-
+0106:         return -p * log(q)
-
    __pyx_r = ((-__pyx_v_p) * log(__pyx_v_q));
-    goto __pyx_L0;
-
 0107:     else:
-
+0108:         return 0.
-
  /*else*/ {
-    __pyx_r = 0.;
-    goto __pyx_L0;
-  }
-
 0109:
-
 0110:
-
+0111: class DecisionTree:
-
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_DecisionTree, __pyx_n_s_DecisionTree, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Tree_Node_Class_Tree_node_class); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 111, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_DecisionTree, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 111, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_DecisionTree, __pyx_t_2) < 0) __PYX_ERR(0, 111, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0112:     """ Tree Node Class
-
 0113:
-
 0114:     Tree node class to contain all the statistics of the tree node.
-
 0115:
-
 0116:     Parameters
-
 0117:     ----------
-
 0118:     classes_ : list of str
-
 0119:         A list of the control and treatment group names.
-
 0120:
-
 0121:     col : int, optional (default = -1)
-
 0122:         The column index for splitting the tree node to children nodes.
-
 0123:
-
 0124:     value : float, optional (default = None)
-
 0125:         The value of the feature column to split the tree node to children nodes.
-
 0126:
-
 0127:     trueBranch : object of DecisionTree
-
 0128:         The true branch tree node (feature > value).
-
 0129:
-
 0130:     falseBranch : object of DecisionTree
-
 0131:         The false branch tree node (feature > value).
-
 0132:
-
 0133:     results : list of float
-
 0134:         The classification probability P(Y=1|T) for each of the control and treatment groups
-
 0135:         in the tree node.
-
 0136:
-
 0137:     summary : list of list
-
 0138:         Summary statistics of the tree nodes, including impurity, sample size, uplift score, etc.
-
 0139:
-
 0140:     maxDiffTreatment : int
-
 0141:         The treatment index generating the maximum difference between the treatment and control groups.
-
 0142:
-
 0143:     maxDiffSign : float
-
 0144:         The sign of the maximum difference (1. or -1.).
-
 0145:
-
 0146:     nodeSummary : list of list
-
 0147:         Summary statistics of the tree nodes [P(Y=1|T), N(T)], where y_mean stands for the target metric mean
-
 0148:         and n is the sample size.
-
 0149:
-
 0150:     backupResults : list of float
-
 0151:         The positive probabilities in each of the control and treatment groups in the parent node. The parent node
-
 0152:         information is served as a backup for the children node, in case no valid statistics can be calculated from the
-
 0153:         children node, the parent node information will be used in certain cases.
-
 0154:
-
 0155:     bestTreatment : int
-
 0156:         The treatment index providing the best uplift (treatment effect).
-
 0157:
-
 0158:     upliftScore : list
-
 0159:         The uplift score of this node: [max_Diff, p_value], where max_Diff stands for the maximum treatment effect, and
-
 0160:         p_value stands for the p_value of the treatment effect.
-
 0161:
-
 0162:     matchScore : float
-
 0163:         The uplift score by filling a trained tree with validation dataset or testing dataset.
-
 0164:
-
 0165:     """
-
 0166:
-
+0167:     def __init__(self, classes_, col=-1, value=None, trueBranch=None, falseBranch=None, results=None, summary=None,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__, METH_VARARGS|METH_KEYWORDS, 0};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_classes_ = 0;
-  PyObject *__pyx_v_col = 0;
-  PyObject *__pyx_v_value = 0;
-  PyObject *__pyx_v_trueBranch = 0;
-  PyObject *__pyx_v_falseBranch = 0;
-  PyObject *__pyx_v_results = 0;
-  PyObject *__pyx_v_summary = 0;
-  PyObject *__pyx_v_maxDiffTreatment = 0;
-  PyObject *__pyx_v_maxDiffSign = 0;
-  PyObject *__pyx_v_nodeSummary = 0;
-  PyObject *__pyx_v_backupResults = 0;
-  PyObject *__pyx_v_bestTreatment = 0;
-  PyObject *__pyx_v_upliftScore = 0;
-  PyObject *__pyx_v_matchScore = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_classes,&__pyx_n_s_col,&__pyx_n_s_value,&__pyx_n_s_trueBranch,&__pyx_n_s_falseBranch,&__pyx_n_s_results,&__pyx_n_s_summary,&__pyx_n_s_maxDiffTreatment,&__pyx_n_s_maxDiffSign,&__pyx_n_s_nodeSummary,&__pyx_n_s_backupResults,&__pyx_n_s_bestTreatment,&__pyx_n_s_upliftScore,&__pyx_n_s_matchScore,0};
-    PyObject* values[15] = {0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
-    values[2] = ((PyObject *)((PyObject *)__pyx_int_neg_1));
-    values[3] = ((PyObject *)((PyObject *)Py_None));
-    values[4] = ((PyObject *)((PyObject *)Py_None));
-    values[5] = ((PyObject *)((PyObject *)Py_None));
-    values[6] = ((PyObject *)((PyObject *)Py_None));
-    values[7] = ((PyObject *)((PyObject *)Py_None));
-/* … */
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_12DecisionTree___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_classes_, PyObject *__pyx_v_col, PyObject *__pyx_v_value, PyObject *__pyx_v_trueBranch, PyObject *__pyx_v_falseBranch, PyObject *__pyx_v_results, PyObject *__pyx_v_summary, PyObject *__pyx_v_maxDiffTreatment, PyObject *__pyx_v_maxDiffSign, PyObject *__pyx_v_nodeSummary, PyObject *__pyx_v_backupResults, PyObject *__pyx_v_bestTreatment, PyObject *__pyx_v_upliftScore, PyObject *__pyx_v_matchScore) {
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__", 0);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.DecisionTree.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__16 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_classes, __pyx_n_s_col, __pyx_n_s_value, __pyx_n_s_trueBranch, __pyx_n_s_falseBranch, __pyx_n_s_results, __pyx_n_s_summary, __pyx_n_s_maxDiffTreatment, __pyx_n_s_maxDiffSign, __pyx_n_s_nodeSummary, __pyx_n_s_backupResults, __pyx_n_s_bestTreatment, __pyx_n_s_upliftScore, __pyx_n_s_matchScore); if (unlikely(!__pyx_tuple__16)) __PYX_ERR(0, 167, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__16);
-  __Pyx_GIVEREF(__pyx_tuple__16);
-  __pyx_codeobj__17 = (PyObject*)__Pyx_PyCode_New(15, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__16, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 167, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__17)) __PYX_ERR(0, 167, __pyx_L1_error)
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_12DecisionTree_1__init__, 0, __pyx_n_s_DecisionTree___init, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__17)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 167, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__18);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_2) < 0) __PYX_ERR(0, 167, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_tuple__18 = PyTuple_Pack(13, ((PyObject *)__pyx_int_neg_1), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject*)__pyx_float_1_), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__18)) __PYX_ERR(0, 167, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__18);
-  __Pyx_GIVEREF(__pyx_tuple__18);
-
+0168:                   maxDiffTreatment=None, maxDiffSign=1., nodeSummary=None, backupResults=None, bestTreatment=None,
-
    values[8] = ((PyObject *)((PyObject *)Py_None));
-    values[9] = ((PyObject *)((PyObject*)__pyx_float_1_));
-    values[10] = ((PyObject *)((PyObject *)Py_None));
-    values[11] = ((PyObject *)((PyObject *)Py_None));
-    values[12] = ((PyObject *)((PyObject *)Py_None));
-
+0169:                   upliftScore=None, matchScore=None):
-
    values[13] = ((PyObject *)((PyObject *)Py_None));
-    values[14] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case 15: values[14] = PyTuple_GET_ITEM(__pyx_args, 14);
-        CYTHON_FALLTHROUGH;
-        case 14: values[13] = PyTuple_GET_ITEM(__pyx_args, 13);
-        CYTHON_FALLTHROUGH;
-        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
-        CYTHON_FALLTHROUGH;
-        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
-        CYTHON_FALLTHROUGH;
-        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
-        CYTHON_FALLTHROUGH;
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_classes)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 15, 1); __PYX_ERR(0, 167, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_col);
-          if (value) { values[2] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_value);
-          if (value) { values[3] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_trueBranch);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_falseBranch);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_results);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_summary);
-          if (value) { values[7] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  8:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_maxDiffTreatment);
-          if (value) { values[8] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  9:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_maxDiffSign);
-          if (value) { values[9] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 10:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_nodeSummary);
-          if (value) { values[10] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 11:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_backupResults);
-          if (value) { values[11] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 12:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_bestTreatment);
-          if (value) { values[12] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 13:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_upliftScore);
-          if (value) { values[13] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 14:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_matchScore);
-          if (value) { values[14] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 167, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case 15: values[14] = PyTuple_GET_ITEM(__pyx_args, 14);
-        CYTHON_FALLTHROUGH;
-        case 14: values[13] = PyTuple_GET_ITEM(__pyx_args, 13);
-        CYTHON_FALLTHROUGH;
-        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
-        CYTHON_FALLTHROUGH;
-        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
-        CYTHON_FALLTHROUGH;
-        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
-        CYTHON_FALLTHROUGH;
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_classes_ = values[1];
-    __pyx_v_col = values[2];
-    __pyx_v_value = values[3];
-    __pyx_v_trueBranch = values[4];
-    __pyx_v_falseBranch = values[5];
-    __pyx_v_results = values[6];
-    __pyx_v_summary = values[7];
-    __pyx_v_maxDiffTreatment = values[8];
-    __pyx_v_maxDiffSign = values[9];
-    __pyx_v_nodeSummary = values[10];
-    __pyx_v_backupResults = values[11];
-    __pyx_v_bestTreatment = values[12];
-    __pyx_v_upliftScore = values[13];
-    __pyx_v_matchScore = values[14];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 15, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 167, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.DecisionTree.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_12DecisionTree___init__(__pyx_self, __pyx_v_self, __pyx_v_classes_, __pyx_v_col, __pyx_v_value, __pyx_v_trueBranch, __pyx_v_falseBranch, __pyx_v_results, __pyx_v_summary, __pyx_v_maxDiffTreatment, __pyx_v_maxDiffSign, __pyx_v_nodeSummary, __pyx_v_backupResults, __pyx_v_bestTreatment, __pyx_v_upliftScore, __pyx_v_matchScore);
-
+0170:         self.classes_ = classes_
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_v_classes_) < 0) __PYX_ERR(0, 170, __pyx_L1_error)
-
+0171:         self.col = col
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_col, __pyx_v_col) < 0) __PYX_ERR(0, 171, __pyx_L1_error)
-
+0172:         self.value = value
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_value, __pyx_v_value) < 0) __PYX_ERR(0, 172, __pyx_L1_error)
-
+0173:         self.trueBranch = trueBranch
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_trueBranch, __pyx_v_trueBranch) < 0) __PYX_ERR(0, 173, __pyx_L1_error)
-
+0174:         self.falseBranch = falseBranch
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_falseBranch, __pyx_v_falseBranch) < 0) __PYX_ERR(0, 174, __pyx_L1_error)
-
+0175:         self.results = results  # None for nodes, not None for leaves
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_results, __pyx_v_results) < 0) __PYX_ERR(0, 175, __pyx_L1_error)
-
+0176:         self.summary = summary
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_summary, __pyx_v_summary) < 0) __PYX_ERR(0, 176, __pyx_L1_error)
-
 0177:         # the treatment with max( |p(y|treatment) - p(y|control)| )
-
+0178:         self.maxDiffTreatment = maxDiffTreatment
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 178, __pyx_L1_error)
-
 0179:         # the sign for p(y|maxDiffTreatment) - p(y|control)
-
+0180:         self.maxDiffSign = maxDiffSign
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 180, __pyx_L1_error)
-
+0181:         self.nodeSummary = nodeSummary
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_nodeSummary, __pyx_v_nodeSummary) < 0) __PYX_ERR(0, 181, __pyx_L1_error)
-
+0182:         self.backupResults = backupResults
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_backupResults, __pyx_v_backupResults) < 0) __PYX_ERR(0, 182, __pyx_L1_error)
-
+0183:         self.bestTreatment = bestTreatment
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 183, __pyx_L1_error)
-
+0184:         self.upliftScore = upliftScore
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 184, __pyx_L1_error)
-
 0185:         # match actual treatment for validation and testing
-
+0186:         self.matchScore = matchScore
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_matchScore, __pyx_v_matchScore) < 0) __PYX_ERR(0, 186, __pyx_L1_error)
-
 0187:
-
 0188:
-
 0189: # Uplift Tree Classifier
-
+0190: class UpliftTreeClassifier:
-
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_UpliftTreeClassifier, __pyx_n_s_UpliftTreeClassifier, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Uplift_Tree_Classifier_for_Clas); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 190, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_UpliftTreeClassifier, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 190, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_UpliftTreeClassifier, __pyx_t_2) < 0) __PYX_ERR(0, 190, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0191:     """ Uplift Tree Classifier for Classification Task.
-
 0192:
-
 0193:     A uplift tree classifier estimates the individual treatment effect by modifying the loss function in the
-
 0194:     classification trees.
-
 0195:
-
 0196:     The uplift tree classifier is used in uplift random forest to construct the trees in the forest.
-
 0197:
-
 0198:     Parameters
-
 0199:     ----------
-
 0200:
-
 0201:     evaluationFunction : string
-
 0202:         Choose from one of the models: 'KL', 'ED', 'Chi', 'CTS', 'DDP'.
-
 0203:
-
 0204:     max_features: int, optional (default=None)
-
 0205:         The number of features to consider when looking for the best split.
-
 0206:
-
 0207:     max_depth: int, optional (default=3)
-
 0208:         The maximum depth of the tree.
-
 0209:
-
 0210:     min_samples_leaf: int, optional (default=100)
-
 0211:         The minimum number of samples required to be split at a leaf node.
-
 0212:
-
 0213:     min_samples_treatment: int, optional (default=10)
-
 0214:         The minimum number of samples required of the experiment group to be split at a leaf node.
-
 0215:
-
 0216:     n_reg: int, optional (default=100)
-
 0217:         The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the
-
 0218:         parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
-
 0219:
-
 0220:     control_name: string
-
 0221:         The name of the control group (other experiment groups will be regarded as treatment groups).
-
 0222:
-
 0223:     normalization: boolean, optional (default=True)
-
 0224:         The normalization factor defined in Rzepakowski et al. 2012, correcting for tests with large number of splits
-
 0225:         and imbalanced treatment and control splits.
-
 0226:
-
 0227:     random_state: int, RandomState instance or None (default=None)
-
 0228:         A random seed or `np.random.RandomState` to control randomness in building a tree.
-
 0229:
-
 0230:     """
-
+0231:     def __init__(self, control_name, max_features=None, max_depth=3, min_samples_leaf=100,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__, METH_VARARGS|METH_KEYWORDS, 0};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_control_name = 0;
-  PyObject *__pyx_v_max_features = 0;
-  PyObject *__pyx_v_max_depth = 0;
-  PyObject *__pyx_v_min_samples_leaf = 0;
-  PyObject *__pyx_v_min_samples_treatment = 0;
-  PyObject *__pyx_v_n_reg = 0;
-  PyObject *__pyx_v_evaluationFunction = 0;
-  PyObject *__pyx_v_normalization = 0;
-  PyObject *__pyx_v_random_state = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_control_name,&__pyx_n_s_max_features,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_evaluationFunction,&__pyx_n_s_normalization,&__pyx_n_s_random_state,0};
-    PyObject* values[10] = {0,0,0,0,0,0,0,0,0,0};
-    values[2] = ((PyObject *)((PyObject *)Py_None));
-    values[3] = ((PyObject *)((PyObject *)__pyx_int_3));
-    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
-    values[5] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[6] = ((PyObject *)((PyObject *)__pyx_int_100));
-    values[7] = ((PyObject *)((PyObject*)__pyx_n_s_KL));
-/* … */
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_control_name, PyObject *__pyx_v_max_features, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_evaluationFunction, PyObject *__pyx_v_normalization, PyObject *__pyx_v_random_state) {
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__", 0);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__19 = PyTuple_Pack(10, __pyx_n_s_self, __pyx_n_s_control_name, __pyx_n_s_max_features, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_evaluationFunction, __pyx_n_s_normalization, __pyx_n_s_random_state); if (unlikely(!__pyx_tuple__19)) __PYX_ERR(0, 231, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__19);
-  __Pyx_GIVEREF(__pyx_tuple__19);
-  __pyx_codeobj__20 = (PyObject*)__Pyx_PyCode_New(10, 0, 10, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__19, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 231, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__20)) __PYX_ERR(0, 231, __pyx_L1_error)
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_1__init__, 0, __pyx_n_s_UpliftTreeClassifier___init, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__20)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 231, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__21);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_2) < 0) __PYX_ERR(0, 231, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_tuple__21 = PyTuple_Pack(8, ((PyObject *)Py_None), ((PyObject *)__pyx_int_3), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject*)__pyx_n_s_KL), ((PyObject *)Py_True), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__21)) __PYX_ERR(0, 231, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__21);
-  __Pyx_GIVEREF(__pyx_tuple__21);
-
 0232:                  min_samples_treatment=10, n_reg=100, evaluationFunction='KL',
-
+0233:                  normalization=True, random_state=None):
-
    values[8] = ((PyObject *)((PyObject *)Py_True));
-    values[9] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 10, 1); __PYX_ERR(0, 231, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_features);
-          if (value) { values[2] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
-          if (value) { values[3] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_evaluationFunction);
-          if (value) { values[7] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  8:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_normalization);
-          if (value) { values[8] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  9:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
-          if (value) { values[9] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 231, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_control_name = values[1];
-    __pyx_v_max_features = values[2];
-    __pyx_v_max_depth = values[3];
-    __pyx_v_min_samples_leaf = values[4];
-    __pyx_v_min_samples_treatment = values[5];
-    __pyx_v_n_reg = values[6];
-    __pyx_v_evaluationFunction = values[7];
-    __pyx_v_normalization = values[8];
-    __pyx_v_random_state = values[9];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 10, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 231, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier___init__(__pyx_self, __pyx_v_self, __pyx_v_control_name, __pyx_v_max_features, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_evaluationFunction, __pyx_v_normalization, __pyx_v_random_state);
-
+0234:         self.max_depth = max_depth
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 234, __pyx_L1_error)
-
+0235:         self.min_samples_leaf = min_samples_leaf
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 235, __pyx_L1_error)
-
+0236:         self.min_samples_treatment = min_samples_treatment
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 236, __pyx_L1_error)
-
+0237:         self.n_reg = n_reg
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 237, __pyx_L1_error)
-
+0238:         self.max_features = max_features
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_features, __pyx_v_max_features) < 0) __PYX_ERR(0, 238, __pyx_L1_error)
-
+0239:         if evaluationFunction == 'KL':
-
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_KL, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 239, __pyx_L1_error)
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L3;
-  }
-
+0240:             self.evaluationFunction = self.evaluate_KL
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_KL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 240, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 240, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0241:         elif evaluationFunction == 'ED':
-
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_ED, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 241, __pyx_L1_error)
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L3;
-  }
-
+0242:             self.evaluationFunction = self.evaluate_ED
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_ED); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 242, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 242, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0243:         elif evaluationFunction == 'Chi':
-
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_Chi, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 243, __pyx_L1_error)
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L3;
-  }
-
+0244:             self.evaluationFunction = self.evaluate_Chi
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_Chi); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 244, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 244, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0245:         elif evaluationFunction == 'DDP':
-
  __pyx_t_1 = (__Pyx_PyString_Equals(__pyx_v_evaluationFunction, __pyx_n_s_DDP, Py_EQ)); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 245, __pyx_L1_error)
-  if (__pyx_t_1) {
-/* … */
-    goto __pyx_L3;
-  }
-
+0246:             self.evaluationFunction = self.evaluate_DDP
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 246, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 246, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0247:         else:
-
+0248:             self.evaluationFunction = self.evaluate_CTS
-
  /*else*/ {
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 248, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_t_2) < 0) __PYX_ERR(0, 248, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  }
-  __pyx_L3:;
-
+0249:         self.fitted_uplift_tree = None
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree, Py_None) < 0) __PYX_ERR(0, 249, __pyx_L1_error)
-
 0250:
-
+0251:         assert control_name is not None and isinstance(control_name, str), \
-
  #ifndef CYTHON_WITHOUT_ASSERTIONS
-  if (unlikely(!Py_OptimizeFlag)) {
-    __pyx_t_3 = (__pyx_v_control_name != Py_None);
-    __pyx_t_4 = (__pyx_t_3 != 0);
-    if (__pyx_t_4) {
-    } else {
-      __pyx_t_1 = __pyx_t_4;
-      goto __pyx_L4_bool_binop_done;
-    }
-    __pyx_t_4 = PyString_Check(__pyx_v_control_name); 
-    __pyx_t_3 = (__pyx_t_4 != 0);
-    __pyx_t_1 = __pyx_t_3;
-    __pyx_L4_bool_binop_done:;
-    if (unlikely(!__pyx_t_1)) {
-
+0252:             f"control_group should be string but {control_name} is passed"
-
      __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 252, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_5 = 0;
-      __pyx_t_6 = 127;
-      __Pyx_INCREF(__pyx_kp_u_control_group_should_be_string_b);
-      __pyx_t_5 += 35;
-      __Pyx_GIVEREF(__pyx_kp_u_control_group_should_be_string_b);
-      PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_kp_u_control_group_should_be_string_b);
-      __pyx_t_7 = __Pyx_PyObject_FormatSimple(__pyx_v_control_name, __pyx_empty_unicode); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 252, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __pyx_t_6 = (__Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) > __pyx_t_6) ? __Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) : __pyx_t_6;
-      __pyx_t_5 += __Pyx_PyUnicode_GET_LENGTH(__pyx_t_7);
-      __Pyx_GIVEREF(__pyx_t_7);
-      PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_t_7);
-      __pyx_t_7 = 0;
-      __Pyx_INCREF(__pyx_kp_u_is_passed);
-      __pyx_t_5 += 10;
-      __Pyx_GIVEREF(__pyx_kp_u_is_passed);
-      PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_kp_u_is_passed);
-      __pyx_t_7 = __Pyx_PyUnicode_Join(__pyx_t_2, 3, __pyx_t_5, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 252, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      PyErr_SetObject(PyExc_AssertionError, __pyx_t_7);
-      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      __PYX_ERR(0, 251, __pyx_L1_error)
-    }
-  }
-  #endif
-
+0253:         self.control_name = control_name
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 253, __pyx_L1_error)
-
+0254:         self.classes_ = [self.control_name]
-
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GIVEREF(__pyx_t_7);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_7);
-  __pyx_t_7 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 254, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0255:         self.n_class = 1
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_int_1) < 0) __PYX_ERR(0, 255, __pyx_L1_error)
-
 0256:
-
+0257:         self.normalization = normalization
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_normalization, __pyx_v_normalization) < 0) __PYX_ERR(0, 257, __pyx_L1_error)
-
+0258:         self.random_state = random_state
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 258, __pyx_L1_error)
-
 0259:
-
+0260:     def fit(self, X, treatment, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit[] = " Fit the uplift model.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        self : object\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 260, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 260, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 260, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 260, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 260, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_treatment_groups = NULL;
-  PyObject *__pyx_v_treatment_idx = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_v_col = NULL;
-  PyObject *__pyx_v_imp = NULL;
-  PyObject *__pyx_v_x = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit", 0);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment);
-  __Pyx_INCREF(__pyx_v_y);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_treatment_groups);
-  __Pyx_XDECREF(__pyx_v_treatment_idx);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XDECREF(__pyx_v_col);
-  __Pyx_XDECREF(__pyx_v_imp);
-  __Pyx_XDECREF(__pyx_v_x);
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XDECREF(__pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_v_y);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__22 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_treatment_groups, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr, __pyx_n_s_col, __pyx_n_s_imp, __pyx_n_s_x); if (unlikely(!__pyx_tuple__22)) __PYX_ERR(0, 260, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__22);
-  __Pyx_GIVEREF(__pyx_tuple__22);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_3fit, 0, __pyx_n_s_UpliftTreeClassifier_fit, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__23)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 260, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fit, __pyx_t_2) < 0) __PYX_ERR(0, 260, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_codeobj__23 = (PyObject*)__Pyx_PyCode_New(4, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__22, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fit, 260, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__23)) __PYX_ERR(0, 260, __pyx_L1_error)
-
 0261:         """ Fit the uplift model.
-
 0262:
-
 0263:         Args
-
 0264:         ----
-
 0265:         X : ndarray, shape = [num_samples, num_features]
-
 0266:             An ndarray of the covariates used to train the uplift model.
-
 0267:
-
 0268:         treatment : array-like, shape = [num_samples]
-
 0269:             An array containing the treatment group for each unit.
-
 0270:
-
 0271:         y : array-like, shape = [num_samples]
-
 0272:             An array containing the outcome of interest for each unit.
-
 0273:
-
 0274:         Returns
-
 0275:         -------
-
 0276:         self : object
-
 0277:         """
-
 0278:
-
+0279:         self.random_state_ = check_random_state(self.random_state)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 279, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 279, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 279, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state_2, __pyx_t_1) < 0) __PYX_ERR(0, 279, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0280:
-
+0281:         X, y = check_X_y(X, y)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 281, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_5 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_5 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_4 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_X);
-    __Pyx_GIVEREF(__pyx_v_X);
-    PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_5, __pyx_v_X);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_5, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_4, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
-    PyObject* sequence = __pyx_t_1;
-    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-    if (unlikely(size != 2)) {
-      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-      __PYX_ERR(0, 281, __pyx_L1_error)
-    }
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    if (likely(PyTuple_CheckExact(sequence))) {
-      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
-      __pyx_t_4 = PyTuple_GET_ITEM(sequence, 1); 
-    } else {
-      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
-      __pyx_t_4 = PyList_GET_ITEM(sequence, 1); 
-    }
-    __Pyx_INCREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_t_4);
-    #else
-    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    #endif
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  } else {
-    Py_ssize_t index = -1;
-    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 281, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
-    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_2);
-    index = 1; __pyx_t_4 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_4)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_4);
-    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 281, __pyx_L1_error)
-    __pyx_t_6 = NULL;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    goto __pyx_L4_unpacking_done;
-    __pyx_L3_unpacking_failed:;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_6 = NULL;
-    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-    __PYX_ERR(0, 281, __pyx_L1_error)
-    __pyx_L4_unpacking_done:;
-  }
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
-  __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_4);
-  __pyx_t_4 = 0;
-
+0282:         treatment = np.asarray(treatment)
-
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 282, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 282, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 282, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
-  __pyx_t_1 = 0;
-
+0283:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
-
  #ifndef CYTHON_WITHOUT_ASSERTIONS
-  if (unlikely(!Py_OptimizeFlag)) {
-    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 283, __pyx_L1_error)
-    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 283, __pyx_L1_error)
-    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
-      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
-      __PYX_ERR(0, 283, __pyx_L1_error)
-    }
-  }
-  #endif
-
 0284:
-
 0285:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
-
+0286:         treatment_groups = sorted([x for x in list(set(treatment)) if x != self.control_name])
-
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 286, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = PySet_New(__pyx_v_treatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 286, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_3 = PySequence_List(__pyx_t_4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __pyx_t_3; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  for (;;) {
-    if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_3); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 286, __pyx_L1_error)
-    #else
-    __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    #endif
-    __Pyx_XDECREF_SET(__pyx_v_x, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 286, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_9 = PyObject_RichCompare(__pyx_v_x, __pyx_t_3, Py_NE); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 286, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 286, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    if (__pyx_t_10) {
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_v_x))) __PYX_ERR(0, 286, __pyx_L1_error)
-    }
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_1 = ((PyObject*)__pyx_t_2);
-  __pyx_t_2 = 0;
-  __pyx_t_11 = PyList_Sort(__pyx_t_1); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 286, __pyx_L1_error)
-  __pyx_v_treatment_groups = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
+0287:         self.classes_ = [self.control_name]
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 287, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 287, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GIVEREF(__pyx_t_1);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
-  __pyx_t_1 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 287, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0288:         treatment_idx = np.zeros_like(treatment, dtype=int)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_treatment);
-  __Pyx_GIVEREF(__pyx_v_treatment);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment);
-  __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_dtype, ((PyObject *)(&PyInt_Type))) < 0) __PYX_ERR(0, 288, __pyx_L1_error)
-  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_v_treatment_idx = __pyx_t_9;
-  __pyx_t_9 = 0;
-
+0289:         for i, tr in enumerate(treatment_groups, 1):
-
  __Pyx_INCREF(__pyx_int_1);
-  __pyx_t_9 = __pyx_int_1;
-  __pyx_t_4 = __pyx_v_treatment_groups; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
-  for (;;) {
-    if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_2 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 289, __pyx_L1_error)
-    #else
-    __pyx_t_2 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    #endif
-    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
-    __pyx_t_2 = 0;
-    __Pyx_INCREF(__pyx_t_9);
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_9);
-    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_9, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_9);
-    __pyx_t_9 = __pyx_t_2;
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-
+0290:             self.classes_.append(tr)
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 290, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_11 = __Pyx_PyObject_Append(__pyx_t_2, __pyx_v_tr); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 290, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0291:             treatment_idx[treatment == tr] = i
-
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 291, __pyx_L1_error)
-    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 291, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0292:         self.n_class = len(self.classes_)
-
  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 292, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_8 = PyObject_Length(__pyx_t_9); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 292, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = PyInt_FromSsize_t(__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 292, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_t_9) < 0) __PYX_ERR(0, 292, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-
 0293:
-
+0294:         self.feature_imp_dict = defaultdict(float)
-
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 294, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_4, function);
-    }
-  }
-  __pyx_t_9 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_2, ((PyObject *)(&PyFloat_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_4, ((PyObject *)(&PyFloat_Type)));
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 294, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict, __pyx_t_9) < 0) __PYX_ERR(0, 294, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-
 0295:
-
+0296:         if self.evaluationFunction == self.evaluate_DDP and self.n_class > 2:
-
  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = PyObject_RichCompare(__pyx_t_9, __pyx_t_4, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__pyx_t_12) {
-  } else {
-    __pyx_t_10 = __pyx_t_12;
-    goto __pyx_L11_bool_binop_done;
-  }
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = PyObject_RichCompare(__pyx_t_2, __pyx_int_2, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_12 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_12 < 0)) __PYX_ERR(0, 296, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_10 = __pyx_t_12;
-  __pyx_L11_bool_binop_done:;
-  if (unlikely(__pyx_t_10)) {
-/* … */
-  }
-
+0297:             raise ValueError("The DDP approach can only cope with two class problems, that is two different treatment "
-
    __pyx_t_4 = __Pyx_PyObject_Call(__pyx_builtin_ValueError, __pyx_tuple__2, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 297, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_Raise(__pyx_t_4, 0, 0, 0);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __PYX_ERR(0, 297, __pyx_L1_error)
-/* … */
-  __pyx_tuple__2 = PyTuple_Pack(1, __pyx_kp_s_The_DDP_approach_can_only_cope_w); if (unlikely(!__pyx_tuple__2)) __PYX_ERR(0, 297, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__2);
-  __Pyx_GIVEREF(__pyx_tuple__2);
-
 0298:                              "options (e.g., control vs treatment). Please select another approach or only use a "
-
 0299:                              "dataset which employs two treatment options.")
-
 0300:
-
+0301:         self.fitted_uplift_tree = self.growDecisionTreeFrom(
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 301, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-/* … */
-  __pyx_t_2 = PyTuple_New(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 301, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_v_X);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_2, 2, __pyx_v_y);
-/* … */
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_2, __pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 301, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree, __pyx_t_1) < 0) __PYX_ERR(0, 301, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 0302:             X, treatment_idx, y,
-
+0303:             max_depth=self.max_depth, min_samples_leaf=self.min_samples_leaf,
-
  __pyx_t_9 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_max_depth, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_leaf, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_depth, __pyx_int_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-
+0304:             depth=1, min_samples_treatment=self.min_samples_treatment,
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 304, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0305:             n_reg=self.n_reg, parentNodeSummary=None
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 305, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_t_1) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 303, __pyx_L1_error)
-
 0306:         )
-
 0307:
-
+0308:         self.feature_importances_ = np.zeros(X.shape[1])
-
  __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_zeros); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_9, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_9)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_9);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_9, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_1) < 0) __PYX_ERR(0, 308, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0309:         for col, imp in self.feature_imp_dict.items():
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 309, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_items); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 309, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_4, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_2) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
-    __pyx_t_4 = __pyx_t_1; __Pyx_INCREF(__pyx_t_4); __pyx_t_8 = 0;
-    __pyx_t_13 = NULL;
-  } else {
-    __pyx_t_8 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 309, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_13 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_13)) __PYX_ERR(0, 309, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_13)) {
-      if (likely(PyList_CheckExact(__pyx_t_4))) {
-        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_4)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_1); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 309, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      } else {
-        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_8); __Pyx_INCREF(__pyx_t_1); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 309, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 309, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      }
-    } else {
-      __pyx_t_1 = __pyx_t_13(__pyx_t_4);
-      if (unlikely(!__pyx_t_1)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 309, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_1);
-    }
-    if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
-      PyObject* sequence = __pyx_t_1;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 2)) {
-        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 309, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_9 = PyTuple_GET_ITEM(sequence, 1); 
-      } else {
-        __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_9 = PyList_GET_ITEM(sequence, 1); 
-      }
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_9);
-      #else
-      __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 309, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_9 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 309, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      #endif
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 309, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
-      index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L15_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_2);
-      index = 1; __pyx_t_9 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_9)) goto __pyx_L15_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_9);
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 309, __pyx_L1_error)
-      __pyx_t_6 = NULL;
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      goto __pyx_L16_unpacking_done;
-      __pyx_L15_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_6 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 309, __pyx_L1_error)
-      __pyx_L16_unpacking_done:;
-    }
-    __Pyx_XDECREF_SET(__pyx_v_col, __pyx_t_2);
-    __pyx_t_2 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_imp, __pyx_t_9);
-    __pyx_t_9 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0310:             self.feature_importances_[col] = imp
-
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 310, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    if (unlikely(PyObject_SetItem(__pyx_t_1, __pyx_v_col, __pyx_v_imp) < 0)) __PYX_ERR(0, 310, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0311:         self.feature_importances_ /= self.feature_importances_.sum()  # normalize to add to 1
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_sum); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_9)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_9);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_9) : __Pyx_PyObject_CallNoArg(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyNumber_InPlaceDivide(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_2) < 0) __PYX_ERR(0, 311, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0312:
-
 0313:     # Prune Trees
-
+0314:     def prune(self, X, treatment, y, minGain=0.0001, rule='maxAbsDiff'):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune[] = " Prune the uplift model.\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        minGain : float, optional (default = 0.0001)\n            The minimum gain required to make a tree node split. The children\n            tree branches are trimmed if the actual split gain is less than\n            the minimum gain.\n        rule : string, optional (default = 'maxAbsDiff')\n            The prune rules. Supported values are 'maxAbsDiff' for optimizing\n            the maximum absolute difference, and 'bestUplift' for optimizing\n            the node-size weighted treatment effect.\n        Returns\n        -------\n        self : object\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune = {"prune", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_minGain = 0;
-  PyObject *__pyx_v_rule = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("prune (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_minGain,&__pyx_n_s_rule,0};
-    PyObject* values[6] = {0,0,0,0,0,0};
-    values[4] = ((PyObject *)((PyObject*)__pyx_float_0_0001));
-    values[5] = ((PyObject *)((PyObject*)__pyx_n_s_maxAbsDiff));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 1); __PYX_ERR(0, 314, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 2); __PYX_ERR(0, 314, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, 3); __PYX_ERR(0, 314, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_minGain);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_rule);
-          if (value) { values[5] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "prune") < 0)) __PYX_ERR(0, 314, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_minGain = values[4];
-    __pyx_v_rule = values[5];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("prune", 0, 4, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 314, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.prune", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_minGain, __pyx_v_rule);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_4prune(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_minGain, PyObject *__pyx_v_rule) {
-  PyObject *__pyx_v_treatment_idx = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("prune", 0);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment);
-  __Pyx_INCREF(__pyx_v_y);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.prune", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_treatment_idx);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XDECREF(__pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_v_y);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__24 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_minGain, __pyx_n_s_rule, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__24)) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__24);
-  __Pyx_GIVEREF(__pyx_tuple__24);
-  __pyx_codeobj__25 = (PyObject*)__Pyx_PyCode_New(6, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__24, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_prune, 314, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__25)) __PYX_ERR(0, 314, __pyx_L1_error)
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_5prune, 0, __pyx_n_s_UpliftTreeClassifier_prune, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__25)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__26);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_prune, __pyx_t_2) < 0) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_tuple__26 = PyTuple_Pack(2, ((PyObject*)__pyx_float_0_0001), ((PyObject*)__pyx_n_s_maxAbsDiff)); if (unlikely(!__pyx_tuple__26)) __PYX_ERR(0, 314, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__26);
-  __Pyx_GIVEREF(__pyx_tuple__26);
-
 0315:         """ Prune the uplift model.
-
 0316:         Args
-
 0317:         ----
-
 0318:         X : ndarray, shape = [num_samples, num_features]
-
 0319:             An ndarray of the covariates used to train the uplift model.
-
 0320:         treatment : array-like, shape = [num_samples]
-
 0321:             An array containing the treatment group for each unit.
-
 0322:         y : array-like, shape = [num_samples]
-
 0323:             An array containing the outcome of interest for each unit.
-
 0324:         minGain : float, optional (default = 0.0001)
-
 0325:             The minimum gain required to make a tree node split. The children
-
 0326:             tree branches are trimmed if the actual split gain is less than
-
 0327:             the minimum gain.
-
 0328:         rule : string, optional (default = 'maxAbsDiff')
-
 0329:             The prune rules. Supported values are 'maxAbsDiff' for optimizing
-
 0330:             the maximum absolute difference, and 'bestUplift' for optimizing
-
 0331:             the node-size weighted treatment effect.
-
 0332:         Returns
-
 0333:         -------
-
 0334:         self : object
-
 0335:         """
-
 0336:
-
+0337:         X, y = check_X_y(X, y)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 337, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_X);
-    __Pyx_GIVEREF(__pyx_v_X);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
-    PyObject* sequence = __pyx_t_1;
-    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-    if (unlikely(size != 2)) {
-      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-      __PYX_ERR(0, 337, __pyx_L1_error)
-    }
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    if (likely(PyTuple_CheckExact(sequence))) {
-      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
-      __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
-    } else {
-      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
-      __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
-    }
-    __Pyx_INCREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_t_5);
-    #else
-    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    #endif
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  } else {
-    Py_ssize_t index = -1;
-    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 337, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
-    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_2);
-    index = 1; __pyx_t_5 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_5)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 337, __pyx_L1_error)
-    __pyx_t_6 = NULL;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    goto __pyx_L4_unpacking_done;
-    __pyx_L3_unpacking_failed:;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_6 = NULL;
-    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-    __PYX_ERR(0, 337, __pyx_L1_error)
-    __pyx_L4_unpacking_done:;
-  }
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
-  __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_5);
-  __pyx_t_5 = 0;
-
+0338:         treatment = np.asarray(treatment)
-
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 338, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 338, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 338, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
-  __pyx_t_1 = 0;
-
+0339:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
-
  #ifndef CYTHON_WITHOUT_ASSERTIONS
-  if (unlikely(!Py_OptimizeFlag)) {
-    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 339, __pyx_L1_error)
-    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 339, __pyx_L1_error)
-    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
-      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
-      __PYX_ERR(0, 339, __pyx_L1_error)
-    }
-  }
-  #endif
-
 0340:
-
 0341:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
-
+0342:         treatment_idx = np.zeros_like(treatment)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 342, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 342, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_5);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_5, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_2, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 342, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_treatment_idx = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+0343:         for i, tr in enumerate(self.classes_[1:], 1):
-
  __Pyx_INCREF(__pyx_int_1);
-  __pyx_t_1 = __pyx_int_1;
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 343, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_t_5, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_8 = 0;
-    __pyx_t_9 = NULL;
-  } else {
-    __pyx_t_8 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 343, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_9 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 343, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_9)) {
-      if (likely(PyList_CheckExact(__pyx_t_5))) {
-        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_5)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 343, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 343, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_9(__pyx_t_5);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 343, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
-    __pyx_t_2 = 0;
-    __Pyx_INCREF(__pyx_t_1);
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
-    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 343, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1);
-    __pyx_t_1 = __pyx_t_2;
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-/* … */
-  __pyx_slice__3 = PySlice_New(__pyx_int_1, Py_None, Py_None); if (unlikely(!__pyx_slice__3)) __PYX_ERR(0, 343, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_slice__3);
-  __Pyx_GIVEREF(__pyx_slice__3);
-
+0344:             treatment_idx[treatment == tr] = i
-
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 344, __pyx_L1_error)
-    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 344, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0345:
-
+0346:         self.pruneTree(X, treatment_idx, y,
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 346, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = PyTuple_New(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 346, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_v_X);
-  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_5, 2, __pyx_v_y);
-/* … */
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 346, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0347:                        tree=self.fitted_uplift_tree,
-
  __pyx_t_2 = __Pyx_PyDict_NewPresized(5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 347, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 347, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_tree, __pyx_t_3) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0348:                        rule=rule,
-
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_rule, __pyx_v_rule) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
-
+0349:                        minGain=minGain,
-
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_minGain, __pyx_v_minGain) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
-
+0350:                        n_reg=self.n_reg,
-
  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 350, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_reg, __pyx_t_3) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0351:                        parentNodeSummary=None)
-
  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 347, __pyx_L1_error)
-
+0352:         return self
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_self);
-  __pyx_r = __pyx_v_self;
-  goto __pyx_L0;
-
 0353:
-
+0354:     def pruneTree(self, X, treatment_idx, y, tree, rule='maxAbsDiff', minGain=0.,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree[] = "Prune one single tree node in the uplift model.\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        rule : string, optional (default = 'maxAbsDiff')\n            The prune rules. Supported values are 'maxAbsDiff' for optimizing the maximum absolute difference, and\n            'bestUplift' for optimizing the node-size weighted treatment effect.\n        minGain : float, optional (default = 0.)\n            The minimum gain required to make a tree node split. The children tree branches are trimmed if the actual\n            split gain is less than the minimum gain.\n        n_reg: int, optional (default=0)\n            The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the\n            parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : list of list, optional (default = None)\n            Node summary statistics, [P(Y=1|T), N(T)] of the parent tree node.\n        Returns\n        -------\n        self : object\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree = {"pruneTree", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_v_rule = 0;
-  PyObject *__pyx_v_minGain = 0;
-  PyObject *__pyx_v_n_reg = 0;
-  PyObject *__pyx_v_parentNodeSummary = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("pruneTree (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_tree,&__pyx_n_s_rule,&__pyx_n_s_minGain,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
-    PyObject* values[9] = {0,0,0,0,0,0,0,0,0};
-    values[5] = ((PyObject *)((PyObject*)__pyx_n_s_maxAbsDiff));
-    values[6] = ((PyObject *)((PyObject*)__pyx_float_0_));
-    values[7] = ((PyObject *)((PyObject *)__pyx_int_0));
-/* … */
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_tree, PyObject *__pyx_v_rule, PyObject *__pyx_v_minGain, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
-  PyObject *__pyx_v_currentNodeSummary = NULL;
-  PyObject *__pyx_v_X_l = NULL;
-  PyObject *__pyx_v_X_r = NULL;
-  PyObject *__pyx_v_w_l = NULL;
-  PyObject *__pyx_v_w_r = NULL;
-  PyObject *__pyx_v_y_l = NULL;
-  PyObject *__pyx_v_y_r = NULL;
-  PyObject *__pyx_v_currentScoreD = NULL;
-  PyObject *__pyx_v_trueNodeSummary = NULL;
-  PyObject *__pyx_v_trueScoreD = NULL;
-  PyObject *__pyx_v_falseNodeSummary = NULL;
-  PyObject *__pyx_v_falseScoreD = NULL;
-  PyObject *__pyx_v_gain = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("pruneTree", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_XDECREF(__pyx_t_12);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.pruneTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
-  __Pyx_XDECREF(__pyx_v_X_l);
-  __Pyx_XDECREF(__pyx_v_X_r);
-  __Pyx_XDECREF(__pyx_v_w_l);
-  __Pyx_XDECREF(__pyx_v_w_r);
-  __Pyx_XDECREF(__pyx_v_y_l);
-  __Pyx_XDECREF(__pyx_v_y_r);
-  __Pyx_XDECREF(__pyx_v_currentScoreD);
-  __Pyx_XDECREF(__pyx_v_trueNodeSummary);
-  __Pyx_XDECREF(__pyx_v_trueScoreD);
-  __Pyx_XDECREF(__pyx_v_falseNodeSummary);
-  __Pyx_XDECREF(__pyx_v_falseScoreD);
-  __Pyx_XDECREF(__pyx_v_gain);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__27 = PyTuple_Pack(22, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_rule, __pyx_n_s_minGain, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_currentNodeSummary, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_currentScoreD, __pyx_n_s_trueNodeSummary, __pyx_n_s_trueScoreD, __pyx_n_s_falseNodeSummary, __pyx_n_s_falseScoreD, __pyx_n_s_gain); if (unlikely(!__pyx_tuple__27)) __PYX_ERR(0, 354, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__27);
-  __Pyx_GIVEREF(__pyx_tuple__27);
-  __pyx_codeobj__28 = (PyObject*)__Pyx_PyCode_New(9, 0, 22, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__27, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_pruneTree, 354, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__28)) __PYX_ERR(0, 354, __pyx_L1_error)
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_7pruneTree, 0, __pyx_n_s_UpliftTreeClassifier_pruneTree, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__28)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 354, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_2, __pyx_tuple__29);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_pruneTree, __pyx_t_2) < 0) __PYX_ERR(0, 354, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_tuple__29 = PyTuple_Pack(4, ((PyObject*)__pyx_n_s_maxAbsDiff), ((PyObject*)__pyx_float_0_), ((PyObject *)__pyx_int_0), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__29)) __PYX_ERR(0, 354, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__29);
-  __Pyx_GIVEREF(__pyx_tuple__29);
-
 0355:                   n_reg=0,
-
+0356:                   parentNodeSummary=None):
-
    values[8] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 1); __PYX_ERR(0, 354, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 2); __PYX_ERR(0, 354, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 3); __PYX_ERR(0, 354, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, 4); __PYX_ERR(0, 354, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_rule);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_minGain);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
-          if (value) { values[7] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  8:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
-          if (value) { values[8] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "pruneTree") < 0)) __PYX_ERR(0, 354, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment_idx = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_tree = values[4];
-    __pyx_v_rule = values[5];
-    __pyx_v_minGain = values[6];
-    __pyx_v_n_reg = values[7];
-    __pyx_v_parentNodeSummary = values[8];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("pruneTree", 0, 5, 9, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 354, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.pruneTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_6pruneTree(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_tree, __pyx_v_rule, __pyx_v_minGain, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
-
 0357:         """Prune one single tree node in the uplift model.
-
 0358:         Args
-
 0359:         ----
-
 0360:         X : ndarray, shape = [num_samples, num_features]
-
 0361:             An ndarray of the covariates used to train the uplift model.
-
 0362:         treatment_idx : array-like, shape = [num_samples]
-
 0363:             An array containing the treatment group index for each unit.
-
 0364:         y : array-like, shape = [num_samples]
-
 0365:             An array containing the outcome of interest for each unit.
-
 0366:         rule : string, optional (default = 'maxAbsDiff')
-
 0367:             The prune rules. Supported values are 'maxAbsDiff' for optimizing the maximum absolute difference, and
-
 0368:             'bestUplift' for optimizing the node-size weighted treatment effect.
-
 0369:         minGain : float, optional (default = 0.)
-
 0370:             The minimum gain required to make a tree node split. The children tree branches are trimmed if the actual
-
 0371:             split gain is less than the minimum gain.
-
 0372:         n_reg: int, optional (default=0)
-
 0373:             The regularization parameter defined in Rzepakowski et al. 2012, the weight (in terms of sample size) of the
-
 0374:             parent node influence on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
-
 0375:         parentNodeSummary : list of list, optional (default = None)
-
 0376:             Node summary statistics, [P(Y=1|T), N(T)] of the parent tree node.
-
 0377:         Returns
-
 0378:         -------
-
 0379:         self : object
-
 0380:         """
-
 0381:         # Current Node Summary for Validation Data Set
-
+0382:         currentNodeSummary = self.tree_node_summary(
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 382, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 382, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_y);
-/* … */
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 382, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_v_currentNodeSummary = __pyx_t_4;
-  __pyx_t_4 = 0;
-
+0383:             treatment_idx, y, min_samples_treatment=self.min_samples_treatment,
-
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 383, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 383, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_t_4) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0384:             n_reg=n_reg, parentNodeSummary=parentNodeSummary
-
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, __pyx_v_parentNodeSummary) < 0) __PYX_ERR(0, 383, __pyx_L1_error)
-
 0385:         )
-
+0386:         tree.nodeSummary = currentNodeSummary
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 386, __pyx_L1_error)
-
 0387:         # Divide sets for child nodes
-
+0388:         if (tree.trueBranch is None) or (tree.falseBranch is None):
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 388, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_6 = (__pyx_t_4 == Py_None);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_7 = (__pyx_t_6 != 0);
-  if (!__pyx_t_7) {
-  } else {
-    __pyx_t_5 = __pyx_t_7;
-    goto __pyx_L4_bool_binop_done;
-  }
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 388, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_7 = (__pyx_t_4 == Py_None);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_6 = (__pyx_t_7 != 0);
-  __pyx_t_5 = __pyx_t_6;
-  __pyx_L4_bool_binop_done:;
-  if (__pyx_t_5) {
-/* … */
-  }
-
+0389:             X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, tree.col, tree.value)
-
    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_8 = NULL;
-    __pyx_t_9 = 0;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
-      __pyx_t_8 = PyMethod_GET_SELF(__pyx_t_3);
-      if (likely(__pyx_t_8)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-        __Pyx_INCREF(__pyx_t_8);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_3, function);
-        __pyx_t_9 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_3)) {
-      PyObject *__pyx_temp[6] = {__pyx_t_8, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
-      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_9, 5+__pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
-      PyObject *__pyx_temp[6] = {__pyx_t_8, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
-      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_9, 5+__pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    } else
-    #endif
-    {
-      __pyx_t_10 = PyTuple_New(5+__pyx_t_9); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 389, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      if (__pyx_t_8) {
-        __Pyx_GIVEREF(__pyx_t_8); PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_t_8); __pyx_t_8 = NULL;
-      }
-      __Pyx_INCREF(__pyx_v_X);
-      __Pyx_GIVEREF(__pyx_v_X);
-      PyTuple_SET_ITEM(__pyx_t_10, 0+__pyx_t_9, __pyx_v_X);
-      __Pyx_INCREF(__pyx_v_treatment_idx);
-      __Pyx_GIVEREF(__pyx_v_treatment_idx);
-      PyTuple_SET_ITEM(__pyx_t_10, 1+__pyx_t_9, __pyx_v_treatment_idx);
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_10, 2+__pyx_t_9, __pyx_v_y);
-      __Pyx_GIVEREF(__pyx_t_2);
-      PyTuple_SET_ITEM(__pyx_t_10, 3+__pyx_t_9, __pyx_t_2);
-      __Pyx_GIVEREF(__pyx_t_1);
-      PyTuple_SET_ITEM(__pyx_t_10, 4+__pyx_t_9, __pyx_t_1);
-      __pyx_t_2 = 0;
-      __pyx_t_1 = 0;
-      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_10, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 389, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
-      PyObject* sequence = __pyx_t_4;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 6)) {
-        if (size > 6) __Pyx_RaiseTooManyValuesError(6);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 389, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_10 = PyTuple_GET_ITEM(sequence, 1); 
-        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 2); 
-        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 3); 
-        __pyx_t_8 = PyTuple_GET_ITEM(sequence, 4); 
-        __pyx_t_11 = PyTuple_GET_ITEM(sequence, 5); 
-      } else {
-        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_10 = PyList_GET_ITEM(sequence, 1); 
-        __pyx_t_1 = PyList_GET_ITEM(sequence, 2); 
-        __pyx_t_2 = PyList_GET_ITEM(sequence, 3); 
-        __pyx_t_8 = PyList_GET_ITEM(sequence, 4); 
-        __pyx_t_11 = PyList_GET_ITEM(sequence, 5); 
-      }
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_10);
-      __Pyx_INCREF(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_8);
-      __Pyx_INCREF(__pyx_t_11);
-      #else
-      {
-        Py_ssize_t i;
-        PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_10,&__pyx_t_1,&__pyx_t_2,&__pyx_t_8,&__pyx_t_11};
-        for (i=0; i < 6; i++) {
-          PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 389, __pyx_L1_error)
-          __Pyx_GOTREF(item);
-          *(temps[i]) = item;
-        }
-      }
-      #endif
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_10,&__pyx_t_1,&__pyx_t_2,&__pyx_t_8,&__pyx_t_11};
-      __pyx_t_12 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 389, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_13 = Py_TYPE(__pyx_t_12)->tp_iternext;
-      for (index=0; index < 6; index++) {
-        PyObject* item = __pyx_t_13(__pyx_t_12); if (unlikely(!item)) goto __pyx_L6_unpacking_failed;
-        __Pyx_GOTREF(item);
-        *(temps[index]) = item;
-      }
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_13(__pyx_t_12), 6) < 0) __PYX_ERR(0, 389, __pyx_L1_error)
-      __pyx_t_13 = NULL;
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      goto __pyx_L7_unpacking_done;
-      __pyx_L6_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      __pyx_t_13 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 389, __pyx_L1_error)
-      __pyx_L7_unpacking_done:;
-    }
-    __pyx_v_X_l = __pyx_t_3;
-    __pyx_t_3 = 0;
-    __pyx_v_X_r = __pyx_t_10;
-    __pyx_t_10 = 0;
-    __pyx_v_w_l = __pyx_t_1;
-    __pyx_t_1 = 0;
-    __pyx_v_w_r = __pyx_t_2;
-    __pyx_t_2 = 0;
-    __pyx_v_y_l = __pyx_t_8;
-    __pyx_t_8 = 0;
-    __pyx_v_y_r = __pyx_t_11;
-    __pyx_t_11 = 0;
-
 0390:
-
 0391:             # recursive call for each branch
-
+0392:             if tree.trueBranch.results is None:
-
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_results); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_11);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_5 = (__pyx_t_11 == Py_None);
-    __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-    __pyx_t_6 = (__pyx_t_5 != 0);
-    if (__pyx_t_6) {
-/* … */
-    }
-
+0393:                 self.pruneTree(X_l, w_l, y_l, tree.trueBranch, rule, minGain,
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 393, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 393, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-/* … */
-      __pyx_t_8 = PyTuple_New(7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 393, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_INCREF(__pyx_v_X_l);
-      __Pyx_GIVEREF(__pyx_v_X_l);
-      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_X_l);
-      __Pyx_INCREF(__pyx_v_w_l);
-      __Pyx_GIVEREF(__pyx_v_w_l);
-      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_w_l);
-      __Pyx_INCREF(__pyx_v_y_l);
-      __Pyx_GIVEREF(__pyx_v_y_l);
-      PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_v_y_l);
-      __Pyx_GIVEREF(__pyx_t_4);
-      PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_4);
-      __Pyx_INCREF(__pyx_v_rule);
-      __Pyx_GIVEREF(__pyx_v_rule);
-      PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_v_rule);
-      __Pyx_INCREF(__pyx_v_minGain);
-      __Pyx_GIVEREF(__pyx_v_minGain);
-      PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_v_minGain);
-      __Pyx_INCREF(__pyx_v_n_reg);
-      __Pyx_GIVEREF(__pyx_v_n_reg);
-      PyTuple_SET_ITEM(__pyx_t_8, 6, __pyx_v_n_reg);
-      __pyx_t_4 = 0;
-/* … */
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_8, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 393, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0394:                                n_reg,
-
+0395:                                parentNodeSummary=currentNodeSummary)
-
      __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 395, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 395, __pyx_L1_error)
-
+0396:             if tree.falseBranch.results is None:
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 396, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_results); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 396, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_6 = (__pyx_t_4 == Py_None);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_5 = (__pyx_t_6 != 0);
-    if (__pyx_t_5) {
-/* … */
-    }
-
+0397:                 self.pruneTree(X_r, w_r, y_r, tree.falseBranch, rule, minGain,
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_pruneTree); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 397, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 397, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-/* … */
-      __pyx_t_8 = PyTuple_New(7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 397, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_INCREF(__pyx_v_X_r);
-      __Pyx_GIVEREF(__pyx_v_X_r);
-      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_X_r);
-      __Pyx_INCREF(__pyx_v_w_r);
-      __Pyx_GIVEREF(__pyx_v_w_r);
-      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_w_r);
-      __Pyx_INCREF(__pyx_v_y_r);
-      __Pyx_GIVEREF(__pyx_v_y_r);
-      PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_v_y_r);
-      __Pyx_GIVEREF(__pyx_t_2);
-      PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_2);
-      __Pyx_INCREF(__pyx_v_rule);
-      __Pyx_GIVEREF(__pyx_v_rule);
-      PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_v_rule);
-      __Pyx_INCREF(__pyx_v_minGain);
-      __Pyx_GIVEREF(__pyx_v_minGain);
-      PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_v_minGain);
-      __Pyx_INCREF(__pyx_v_n_reg);
-      __Pyx_GIVEREF(__pyx_v_n_reg);
-      PyTuple_SET_ITEM(__pyx_t_8, 6, __pyx_v_n_reg);
-      __pyx_t_2 = 0;
-/* … */
-      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 397, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
 0398:                                n_reg,
-
+0399:                                parentNodeSummary=currentNodeSummary)
-
      __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 399, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 399, __pyx_L1_error)
-
 0400:
-
 0401:         # merge leaves (potentially)
-
+0402:         if (tree.trueBranch.results is not None and
-
  __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_results); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 402, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __pyx_t_6 = (__pyx_t_2 != Py_None);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_7 = (__pyx_t_6 != 0);
-  if (__pyx_t_7) {
-  } else {
-    __pyx_t_5 = __pyx_t_7;
-    goto __pyx_L11_bool_binop_done;
-  }
-/* … */
-  if (__pyx_t_5) {
-/* … */
-  }
-
+0403:             tree.falseBranch.results is not None):
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 403, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_results); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 403, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_7 = (__pyx_t_11 != Py_None);
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __pyx_t_6 = (__pyx_t_7 != 0);
-  __pyx_t_5 = __pyx_t_6;
-  __pyx_L11_bool_binop_done:;
-
+0404:             if rule == 'maxAbsDiff':
-
    __pyx_t_5 = (__Pyx_PyString_Equals(__pyx_v_rule, __pyx_n_s_maxAbsDiff, Py_EQ)); if (unlikely(__pyx_t_5 < 0)) __PYX_ERR(0, 404, __pyx_L1_error)
-    if (__pyx_t_5) {
-/* … */
-      goto __pyx_L13;
-    }
-
 0405:                 # Current D
-
+0406:                 if (tree.maxDiffTreatment in currentNodeSummary and
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 406, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 406, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L15_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L14;
-      }
-
+0407:                     self.control_name in currentNodeSummary):
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 407, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 407, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L15_bool_binop_done:;
-
+0408:                     currentScoreD = tree.maxDiffSign * (currentNodeSummary[tree.maxDiffTreatment][0]
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 408, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 408, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 408, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 408, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-/* … */
-        __pyx_t_8 = PyNumber_Multiply(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 408, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_v_currentScoreD = __pyx_t_8;
-        __pyx_t_8 = 0;
-
+0409:                                                         - currentNodeSummary[self.control_name][0])
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 409, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 409, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 409, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = PyNumber_Subtract(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 409, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
 0410:                 else:
-
+0411:                     currentScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_currentScoreD = __pyx_int_0;
-      }
-      __pyx_L14:;
-
 0412:
-
 0413:                 # trueBranch D
-
+0414:                 trueNodeSummary = self.tree_node_summary(
-
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 414, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-/* … */
-      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 414, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_INCREF(__pyx_v_w_l);
-      __Pyx_GIVEREF(__pyx_v_w_l);
-      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_l);
-      __Pyx_INCREF(__pyx_v_y_l);
-      __Pyx_GIVEREF(__pyx_v_y_l);
-      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_l);
-/* … */
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_8, __pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 414, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_v_trueNodeSummary = __pyx_t_2;
-      __pyx_t_2 = 0;
-
+0415:                     w_l, y_l, min_samples_treatment=self.min_samples_treatment,
-
      if (unlikely(!__pyx_v_w_l)) { __Pyx_RaiseUnboundLocalError("w_l"); __PYX_ERR(0, 415, __pyx_L1_error) }
-      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 415, __pyx_L1_error) }
-/* … */
-      __pyx_t_11 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 415, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 415, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_min_samples_treatment, __pyx_t_2) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0416:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 415, __pyx_L1_error)
-
 0417:                 )
-
+0418:                 if (tree.trueBranch.maxDiffTreatment in trueNodeSummary and
-
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 418, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 418, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 418, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L18_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L17;
-      }
-
+0419:                     self.control_name in trueNodeSummary):
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 419, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 419, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L18_bool_binop_done:;
-
+0420:                     trueScoreD = tree.trueBranch.maxDiffSign * (trueNodeSummary[tree.trueBranch.maxDiffTreatment][0]
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-/* … */
-        __pyx_t_11 = PyNumber_Multiply(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 420, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_v_trueScoreD = __pyx_t_11;
-        __pyx_t_11 = 0;
-
+0421:                                                                 - trueNodeSummary[self.control_name][0])
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 421, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 421, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 421, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = PyNumber_Subtract(__pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 421, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
 0422:                     trueScoreD = (
-
 0423:                         trueScoreD
-
+0424:                         * (trueNodeSummary[tree.trueBranch.maxDiffTreatment][1]
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 424, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_11, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 424, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-/* … */
-        __pyx_t_11 = PyNumber_Multiply(__pyx_v_trueScoreD, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 424, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0425:                          + trueNodeSummary[self.control_name][1])
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 425, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 425, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 425, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = PyNumber_Add(__pyx_t_8, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 425, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0426:                         / (currentNodeSummary[tree.trueBranch.maxDiffTreatment][1]
-
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 426, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_2, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 426, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-/* … */
-        __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 426, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_DECREF_SET(__pyx_v_trueScoreD, __pyx_t_2);
-        __pyx_t_2 = 0;
-
+0427:                            + currentNodeSummary[self.control_name][1])
-
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 427, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 427, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 427, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = PyNumber_Add(__pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 427, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0428:                     )
-
 0429:                 else:
-
+0430:                     trueScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_trueScoreD = __pyx_int_0;
-      }
-      __pyx_L17:;
-
 0431:
-
 0432:                 # falseBranch D
-
+0433:                 falseNodeSummary = self.tree_node_summary(
-
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 433, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-/* … */
-      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 433, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_INCREF(__pyx_v_w_r);
-      __Pyx_GIVEREF(__pyx_v_w_r);
-      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_r);
-      __Pyx_INCREF(__pyx_v_y_r);
-      __Pyx_GIVEREF(__pyx_v_y_r);
-      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_r);
-/* … */
-      __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 433, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_v_falseNodeSummary = __pyx_t_8;
-      __pyx_t_8 = 0;
-
+0434:                     w_r, y_r, min_samples_treatment=self.min_samples_treatment,
-
      if (unlikely(!__pyx_v_w_r)) { __Pyx_RaiseUnboundLocalError("w_r"); __PYX_ERR(0, 434, __pyx_L1_error) }
-      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 434, __pyx_L1_error) }
-/* … */
-      __pyx_t_11 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 434, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 434, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_min_samples_treatment, __pyx_t_8) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+0435:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_11, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 434, __pyx_L1_error)
-
 0436:                 )
-
+0437:                 if (tree.falseBranch.maxDiffTreatment in falseNodeSummary and
-
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 437, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 437, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 437, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L21_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L20;
-      }
-
+0438:                     self.control_name in falseNodeSummary):
-
-      /* "causalml/inference/tree/uplift.pyx":438
- *                 )
- *                 if (tree.falseBranch.maxDiffTreatment in falseNodeSummary and
- *                     self.control_name in falseNodeSummary):             # <<<<<<<<<<<<<<
- *                     falseScoreD = (
- *                         tree.falseBranch.maxDiffSign *
- */
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 438, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_11, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 438, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L21_bool_binop_done:;
-
 0439:                     falseScoreD = (
-
+0440:                         tree.falseBranch.maxDiffSign *
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 440, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffSign); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 440, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-/* … */
-        __pyx_t_11 = PyNumber_Multiply(__pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 440, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_v_falseScoreD = __pyx_t_11;
-        __pyx_t_11 = 0;
-
+0441:                         (falseNodeSummary[tree.falseBranch.maxDiffTreatment][0]
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 441, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 441, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 441, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 441, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0442:                          - falseNodeSummary[self.control_name][0])
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 442, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 442, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 442, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = PyNumber_Subtract(__pyx_t_4, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 442, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
 0443:                     )
-
 0444:
-
 0445:                     falseScoreD = (
-
+0446:                         falseScoreD *
-
        __pyx_t_11 = PyNumber_Multiply(__pyx_v_falseScoreD, __pyx_t_8); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 446, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+0447:                         (falseNodeSummary[tree.falseBranch.maxDiffTreatment][1]
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 447, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 447, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 447, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_11, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 447, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0448:                          + falseNodeSummary[self.control_name][1])
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 448, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 448, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_8, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 448, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = PyNumber_Add(__pyx_t_2, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 448, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0449:                         / (currentNodeSummary[tree.falseBranch.maxDiffTreatment][1]
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_maxDiffTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 449, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 449, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-/* … */
-        __pyx_t_8 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 449, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_DECREF_SET(__pyx_v_falseScoreD, __pyx_t_8);
-        __pyx_t_8 = 0;
-
+0450:                            + currentNodeSummary[self.control_name][1])
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 450, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 450, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 450, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = PyNumber_Add(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 450, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
 0451:                     )
-
 0452:                 else:
-
+0453:                     falseScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_falseScoreD = __pyx_int_0;
-      }
-      __pyx_L20:;
-
 0454:
-
+0455:                 if ((trueScoreD + falseScoreD) - currentScoreD <= minGain or
-
      __pyx_t_8 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 455, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_4 = PyNumber_Subtract(__pyx_t_8, __pyx_v_currentScoreD); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 455, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_8 = PyObject_RichCompare(__pyx_t_4, __pyx_v_minGain, Py_LE); __Pyx_XGOTREF(__pyx_t_8); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 455, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_8); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 455, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      if (!__pyx_t_6) {
-      } else {
-        __pyx_t_5 = __pyx_t_6;
-        goto __pyx_L24_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-      }
-
+0456:                     (trueScoreD + falseScoreD < 0.)):
-
      __pyx_t_8 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 456, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_4 = PyObject_RichCompare(__pyx_t_8, __pyx_float_0_, Py_LT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 456, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 456, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L24_bool_binop_done:;
-
+0457:                     tree.trueBranch, tree.falseBranch = None, None
-
        __pyx_t_4 = Py_None;
-        __Pyx_INCREF(__pyx_t_4);
-        __pyx_t_8 = Py_None;
-        __Pyx_INCREF(__pyx_t_8);
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch, __pyx_t_4) < 0) __PYX_ERR(0, 457, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch, __pyx_t_8) < 0) __PYX_ERR(0, 457, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+0458:                     tree.results = tree.backupResults
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_backupResults); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 458, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_8) < 0) __PYX_ERR(0, 458, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
 0459:
-
+0460:             elif rule == 'bestUplift':
-
    __pyx_t_5 = (__Pyx_PyString_Equals(__pyx_v_rule, __pyx_n_s_bestUplift, Py_EQ)); if (unlikely(__pyx_t_5 < 0)) __PYX_ERR(0, 460, __pyx_L1_error)
-    if (__pyx_t_5) {
-/* … */
-    }
-    __pyx_L13:;
-
 0461:                 # Current D
-
+0462:                 if (tree.bestTreatment in currentNodeSummary and
-
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 462, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 462, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L27_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L26;
-      }
-
+0463:                     self.control_name in currentNodeSummary):
-
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 463, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_currentNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 463, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L27_bool_binop_done:;
-
 0464:                     currentScoreD = (
-
+0465:                         currentNodeSummary[tree.bestTreatment][0]
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 465, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 465, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 465, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0466:                         - currentNodeSummary[self.control_name][0]
-
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 466, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __pyx_t_11 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 466, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_11, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 466, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = PyNumber_Subtract(__pyx_t_8, __pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 466, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_v_currentScoreD = __pyx_t_11;
-        __pyx_t_11 = 0;
-
 0467:                     )
-
 0468:                 else:
-
+0469:                     currentScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_currentScoreD = __pyx_int_0;
-      }
-      __pyx_L26:;
-
 0470:
-
 0471:                 # trueBranch D
-
+0472:                 trueNodeSummary = self.tree_node_summary(
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 472, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-/* … */
-      __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 472, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_INCREF(__pyx_v_w_l);
-      __Pyx_GIVEREF(__pyx_v_w_l);
-      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_w_l);
-      __Pyx_INCREF(__pyx_v_y_l);
-      __Pyx_GIVEREF(__pyx_v_y_l);
-      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y_l);
-/* … */
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_4, __pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 472, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_v_trueNodeSummary = __pyx_t_2;
-      __pyx_t_2 = 0;
-
+0473:                     w_l, y_l, min_samples_treatment=self.min_samples_treatment,
-
      if (unlikely(!__pyx_v_w_l)) { __Pyx_RaiseUnboundLocalError("w_l"); __PYX_ERR(0, 473, __pyx_L1_error) }
-      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 473, __pyx_L1_error) }
-/* … */
-      __pyx_t_8 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 473, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 473, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_min_samples_treatment, __pyx_t_2) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0474:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 473, __pyx_L1_error)
-
 0475:                 )
-
+0476:                 if (tree.trueBranch.bestTreatment in trueNodeSummary and
-
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 476, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 476, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 476, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L30_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L29;
-      }
-
+0477:                     self.control_name in trueNodeSummary):
-
      __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 477, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_8, __pyx_v_trueNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 477, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L30_bool_binop_done:;
-
 0478:                     trueScoreD = (
-
+0479:                         trueNodeSummary[tree.trueBranch.bestTreatment][0]
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 479, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 479, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 479, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 479, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+0480:                         - trueNodeSummary[self.control_name][0]
-
        __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 480, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_trueNodeSummary, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 480, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 480, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_4 = PyNumber_Subtract(__pyx_t_2, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 480, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_v_trueScoreD = __pyx_t_4;
-        __pyx_t_4 = 0;
-
 0481:                     )
-
 0482:                 else:
-
+0483:                     trueScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_trueScoreD = __pyx_int_0;
-      }
-      __pyx_L29:;
-
 0484:
-
 0485:                 # falseBranch D
-
+0486:                 falseNodeSummary = self.tree_node_summary(
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 486, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-/* … */
-      __pyx_t_8 = PyTuple_New(2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 486, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_INCREF(__pyx_v_w_r);
-      __Pyx_GIVEREF(__pyx_v_w_r);
-      PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_w_r);
-      __Pyx_INCREF(__pyx_v_y_r);
-      __Pyx_GIVEREF(__pyx_v_y_r);
-      PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_v_y_r);
-/* … */
-      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_2); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 486, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_v_falseNodeSummary = __pyx_t_11;
-      __pyx_t_11 = 0;
-
+0487:                     w_r, y_r, min_samples_treatment=self.min_samples_treatment,
-
      if (unlikely(!__pyx_v_w_r)) { __Pyx_RaiseUnboundLocalError("w_r"); __PYX_ERR(0, 487, __pyx_L1_error) }
-      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 487, __pyx_L1_error) }
-/* … */
-      __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 487, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 487, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_min_samples_treatment, __pyx_t_11) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0488:                     n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 487, __pyx_L1_error)
-
 0489:                 )
-
+0490:                 if (tree.falseBranch.bestTreatment in falseNodeSummary and
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 490, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_11, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 490, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_6 = (__Pyx_PySequence_ContainsTF(__pyx_t_2, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 490, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_7 = (__pyx_t_6 != 0);
-      if (__pyx_t_7) {
-      } else {
-        __pyx_t_5 = __pyx_t_7;
-        goto __pyx_L33_bool_binop_done;
-      }
-/* … */
-      if (__pyx_t_5) {
-/* … */
-        goto __pyx_L32;
-      }
-
+0491:                     self.control_name in falseNodeSummary):
-
      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 491, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_7 = (__Pyx_PySequence_ContainsTF(__pyx_t_2, __pyx_v_falseNodeSummary, Py_EQ)); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 491, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_6 = (__pyx_t_7 != 0);
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L33_bool_binop_done:;
-
 0492:                     falseScoreD = (
-
+0493:                         falseNodeSummary[tree.falseBranch.bestTreatment][0]
-
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 493, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 493, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_11); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 493, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __pyx_t_11 = __Pyx_GetItemInt(__pyx_t_2, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 493, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0494:                         - falseNodeSummary[self.control_name][0]
-
        __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 494, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_falseNodeSummary, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 494, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_8, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 494, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __pyx_t_8 = PyNumber_Subtract(__pyx_t_11, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 494, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __pyx_v_falseScoreD = __pyx_t_8;
-        __pyx_t_8 = 0;
-
 0495:                     )
-
 0496:                 else:
-
+0497:                     falseScoreD = 0
-
      /*else*/ {
-        __Pyx_INCREF(__pyx_int_0);
-        __pyx_v_falseScoreD = __pyx_int_0;
-      }
-      __pyx_L32:;
-
+0498:                 gain = ((1. * len(y_l) / len(y) * trueScoreD
-
      if (unlikely(!__pyx_v_y_l)) { __Pyx_RaiseUnboundLocalError("y_l"); __PYX_ERR(0, 498, __pyx_L1_error) }
-      __pyx_t_14 = PyObject_Length(__pyx_v_y_l); if (unlikely(__pyx_t_14 == ((Py_ssize_t)-1))) __PYX_ERR(0, 498, __pyx_L1_error)
-      __pyx_t_15 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 498, __pyx_L1_error)
-      __pyx_t_8 = PyFloat_FromDouble(((1. * __pyx_t_14) / __pyx_t_15)); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 498, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_2 = PyNumber_Multiply(__pyx_t_8, __pyx_v_trueScoreD); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 498, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+0499:                          + 1. * len(y_r) / len(y) * falseScoreD)
-
      if (unlikely(!__pyx_v_y_r)) { __Pyx_RaiseUnboundLocalError("y_r"); __PYX_ERR(0, 499, __pyx_L1_error) }
-      __pyx_t_15 = PyObject_Length(__pyx_v_y_r); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 499, __pyx_L1_error)
-      __pyx_t_14 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_14 == ((Py_ssize_t)-1))) __PYX_ERR(0, 499, __pyx_L1_error)
-      __pyx_t_8 = PyFloat_FromDouble(((1. * __pyx_t_15) / __pyx_t_14)); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 499, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __pyx_t_11 = PyNumber_Multiply(__pyx_t_8, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 499, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_8 = PyNumber_Add(__pyx_t_2, __pyx_t_11); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 499, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0500:                         - currentScoreD)
-
      __pyx_t_11 = PyNumber_Subtract(__pyx_t_8, __pyx_v_currentScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 500, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_v_gain = __pyx_t_11;
-      __pyx_t_11 = 0;
-
+0501:                 if gain <= minGain or (trueScoreD + falseScoreD < 0.):
-
      __pyx_t_11 = PyObject_RichCompare(__pyx_v_gain, __pyx_v_minGain, Py_LE); __Pyx_XGOTREF(__pyx_t_11); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 501, __pyx_L1_error)
-      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_11); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 501, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      if (!__pyx_t_6) {
-      } else {
-        __pyx_t_5 = __pyx_t_6;
-        goto __pyx_L36_bool_binop_done;
-      }
-      __pyx_t_11 = PyNumber_Add(__pyx_v_trueScoreD, __pyx_v_falseScoreD); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 501, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_8 = PyObject_RichCompare(__pyx_t_11, __pyx_float_0_, Py_LT); __Pyx_XGOTREF(__pyx_t_8); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 501, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_8); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 501, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-      __pyx_t_5 = __pyx_t_6;
-      __pyx_L36_bool_binop_done:;
-      if (__pyx_t_5) {
-/* … */
-      }
-
+0502:                     tree.trueBranch, tree.falseBranch = None, None
-
        __pyx_t_8 = Py_None;
-        __Pyx_INCREF(__pyx_t_8);
-        __pyx_t_11 = Py_None;
-        __Pyx_INCREF(__pyx_t_11);
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch, __pyx_t_8) < 0) __PYX_ERR(0, 502, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch, __pyx_t_11) < 0) __PYX_ERR(0, 502, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0503:                     tree.results = tree.backupResults
-
        __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_backupResults); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 503, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_11) < 0) __PYX_ERR(0, 503, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-
+0504:         return self
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_self);
-  __pyx_r = __pyx_v_self;
-  goto __pyx_L0;
-
 0505:
-
+0506:     def fill(self, X, treatment, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill[] = " Fill the data into an existing tree.\n        This is a higher-level function to transform the original data inputs\n        into lower level data inputs (list of list and tree).\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        self : object\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill = {"fill", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fill (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 1); __PYX_ERR(0, 506, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 2); __PYX_ERR(0, 506, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, 3); __PYX_ERR(0, 506, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fill") < 0)) __PYX_ERR(0, 506, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fill", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 506, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fill", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8fill(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_treatment_idx = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fill", 0);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment);
-  __Pyx_INCREF(__pyx_v_y);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fill", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_treatment_idx);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XDECREF(__pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_v_y);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__30 = PyTuple_Pack(7, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_treatment_idx, __pyx_n_s_i, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__30)) __PYX_ERR(0, 506, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__30);
-  __Pyx_GIVEREF(__pyx_tuple__30);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_9fill, 0, __pyx_n_s_UpliftTreeClassifier_fill, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__31)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 506, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fill, __pyx_t_2) < 0) __PYX_ERR(0, 506, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_codeobj__31 = (PyObject*)__Pyx_PyCode_New(4, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__30, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fill, 506, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__31)) __PYX_ERR(0, 506, __pyx_L1_error)
-
 0507:         """ Fill the data into an existing tree.
-
 0508:         This is a higher-level function to transform the original data inputs
-
 0509:         into lower level data inputs (list of list and tree).
-
 0510:
-
 0511:         Args
-
 0512:         ----
-
 0513:         X : ndarray, shape = [num_samples, num_features]
-
 0514:             An ndarray of the covariates used to train the uplift model.
-
 0515:         treatment : array-like, shape = [num_samples]
-
 0516:             An array containing the treatment group for each unit.
-
 0517:         y : array-like, shape = [num_samples]
-
 0518:             An array containing the outcome of interest for each unit.
-
 0519:
-
 0520:         Returns
-
 0521:         -------
-
 0522:         self : object
-
 0523:         """
-
 0524:
-
+0525:         X, y = check_X_y(X, y)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_X_y); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 525, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_X, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_X);
-    __Pyx_GIVEREF(__pyx_v_X);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_X);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
-    PyObject* sequence = __pyx_t_1;
-    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-    if (unlikely(size != 2)) {
-      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-      __PYX_ERR(0, 525, __pyx_L1_error)
-    }
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    if (likely(PyTuple_CheckExact(sequence))) {
-      __pyx_t_2 = PyTuple_GET_ITEM(sequence, 0); 
-      __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
-    } else {
-      __pyx_t_2 = PyList_GET_ITEM(sequence, 0); 
-      __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
-    }
-    __Pyx_INCREF(__pyx_t_2);
-    __Pyx_INCREF(__pyx_t_5);
-    #else
-    __pyx_t_2 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    #endif
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  } else {
-    Py_ssize_t index = -1;
-    __pyx_t_3 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 525, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_6 = Py_TYPE(__pyx_t_3)->tp_iternext;
-    index = 0; __pyx_t_2 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_2)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_2);
-    index = 1; __pyx_t_5 = __pyx_t_6(__pyx_t_3); if (unlikely(!__pyx_t_5)) goto __pyx_L3_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__Pyx_IternextUnpackEndCheck(__pyx_t_6(__pyx_t_3), 2) < 0) __PYX_ERR(0, 525, __pyx_L1_error)
-    __pyx_t_6 = NULL;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    goto __pyx_L4_unpacking_done;
-    __pyx_L3_unpacking_failed:;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_6 = NULL;
-    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-    __PYX_ERR(0, 525, __pyx_L1_error)
-    __pyx_L4_unpacking_done:;
-  }
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_2);
-  __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_y, __pyx_t_5);
-  __pyx_t_5 = 0;
-
+0526:         treatment = np.asarray(treatment)
-
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 526, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_asarray); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 526, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_5, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 526, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_treatment, __pyx_t_1);
-  __pyx_t_1 = 0;
-
+0527:         assert len(y) == len(treatment), 'Data length must be equal for X, treatment, and y.'
-
  #ifndef CYTHON_WITHOUT_ASSERTIONS
-  if (unlikely(!Py_OptimizeFlag)) {
-    __pyx_t_7 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_7 == ((Py_ssize_t)-1))) __PYX_ERR(0, 527, __pyx_L1_error)
-    __pyx_t_8 = PyObject_Length(__pyx_v_treatment); if (unlikely(__pyx_t_8 == ((Py_ssize_t)-1))) __PYX_ERR(0, 527, __pyx_L1_error)
-    if (unlikely(!((__pyx_t_7 == __pyx_t_8) != 0))) {
-      PyErr_SetObject(PyExc_AssertionError, __pyx_kp_s_Data_length_must_be_equal_for_X);
-      __PYX_ERR(0, 527, __pyx_L1_error)
-    }
-  }
-  #endif
-
 0528:
-
 0529:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
-
+0530:         treatment_idx = np.zeros_like(treatment)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 530, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_zeros_like); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 530, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_5);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_5, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_2, __pyx_v_treatment) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_v_treatment);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 530, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_treatment_idx = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+0531:         for i, tr in enumerate(self.classes_[1:], 1):
-
  __Pyx_INCREF(__pyx_int_1);
-  __pyx_t_1 = __pyx_int_1;
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 531, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_t_5, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_8 = 0;
-    __pyx_t_9 = NULL;
-  } else {
-    __pyx_t_8 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 531, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_9 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 531, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_9)) {
-      if (likely(PyList_CheckExact(__pyx_t_5))) {
-        if (__pyx_t_8 >= PyList_GET_SIZE(__pyx_t_5)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 531, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_8 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_8); __Pyx_INCREF(__pyx_t_2); __pyx_t_8++; if (unlikely(0 < 0)) __PYX_ERR(0, 531, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_8); __pyx_t_8++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_9(__pyx_t_5);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 531, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_2);
-    __pyx_t_2 = 0;
-    __Pyx_INCREF(__pyx_t_1);
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
-    __pyx_t_2 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 531, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1);
-    __pyx_t_1 = __pyx_t_2;
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0532:             treatment_idx[treatment == tr] = i
-
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment, __pyx_v_tr, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 532, __pyx_L1_error)
-    if (unlikely(PyObject_SetItem(__pyx_v_treatment_idx, __pyx_t_2, __pyx_v_i) < 0)) __PYX_ERR(0, 532, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0533:
-
+0534:         self.fillTree(X, treatment_idx, y, tree=self.fitted_uplift_tree)
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_5 = PyTuple_New(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_INCREF(__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_v_X);
-  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_v_X);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_5, 2, __pyx_v_y);
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_tree, __pyx_t_3) < 0) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 534, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0535:         return self
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_self);
-  __pyx_r = __pyx_v_self;
-  goto __pyx_L0;
-
 0536:
-
+0537:     def fillTree(self, X, treatment_idx, y, tree):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree[] = " Fill the data into an existing tree.\n        This is a lower-level function to execute on the tree filling task.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        tree : object\n            object of DecisionTree class\n\n        Returns\n        -------\n        self : object\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree = {"fillTree", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fillTree (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_tree,0};
-    PyObject* values[5] = {0,0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 1); __PYX_ERR(0, 537, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 2); __PYX_ERR(0, 537, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 3); __PYX_ERR(0, 537, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, 4); __PYX_ERR(0, 537, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fillTree") < 0)) __PYX_ERR(0, 537, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 5) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-      values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment_idx = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_tree = values[4];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fillTree", 1, 5, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 537, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fillTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_tree);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_10fillTree(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_tree) {
-  PyObject *__pyx_v_currentNodeSummary = NULL;
-  PyObject *__pyx_v_X_l = NULL;
-  PyObject *__pyx_v_X_r = NULL;
-  PyObject *__pyx_v_w_l = NULL;
-  PyObject *__pyx_v_w_r = NULL;
-  PyObject *__pyx_v_y_l = NULL;
-  PyObject *__pyx_v_y_r = NULL;
-  PyObject *__pyx_v_matchScore = NULL;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_v_summary = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fillTree", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_XDECREF(__pyx_t_15);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.fillTree", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
-  __Pyx_XDECREF(__pyx_v_X_l);
-  __Pyx_XDECREF(__pyx_v_X_r);
-  __Pyx_XDECREF(__pyx_v_w_l);
-  __Pyx_XDECREF(__pyx_v_w_r);
-  __Pyx_XDECREF(__pyx_v_y_l);
-  __Pyx_XDECREF(__pyx_v_y_r);
-  __Pyx_XDECREF(__pyx_v_matchScore);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XDECREF(__pyx_v_summary);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__32 = PyTuple_Pack(15, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_currentNodeSummary, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_matchScore, __pyx_n_s_treatment_group, __pyx_n_s_summary); if (unlikely(!__pyx_tuple__32)) __PYX_ERR(0, 537, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__32);
-  __Pyx_GIVEREF(__pyx_tuple__32);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_11fillTree, 0, __pyx_n_s_UpliftTreeClassifier_fillTree, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__33)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 537, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fillTree, __pyx_t_2) < 0) __PYX_ERR(0, 537, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_codeobj__33 = (PyObject*)__Pyx_PyCode_New(5, 0, 15, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__32, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fillTree, 537, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__33)) __PYX_ERR(0, 537, __pyx_L1_error)
-
 0538:         """ Fill the data into an existing tree.
-
 0539:         This is a lower-level function to execute on the tree filling task.
-
 0540:
-
 0541:         Args
-
 0542:         ----
-
 0543:         X : ndarray, shape = [num_samples, num_features]
-
 0544:             An ndarray of the covariates used to train the uplift model.
-
 0545:         treatment_idx : array-like, shape = [num_samples]
-
 0546:             An array containing the treatment group index for each unit.
-
 0547:         y : array-like, shape = [num_samples]
-
 0548:             An array containing the outcome of interest for each unit.
-
 0549:         tree : object
-
 0550:             object of DecisionTree class
-
 0551:
-
 0552:         Returns
-
 0553:         -------
-
 0554:         self : object
-
 0555:         """
-
 0556:         # Current Node Summary for Validation Data Set
-
+0557:         currentNodeSummary = self.tree_node_summary(treatment_idx, y,
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 557, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 557, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_y);
-/* … */
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 557, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_v_currentNodeSummary = __pyx_t_4;
-  __pyx_t_4 = 0;
-
+0558:                                                     min_samples_treatment=0,
-
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 558, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_int_0) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_int_0) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
-
 0559:                                                     n_reg=0,
-
+0560:                                                     parentNodeSummary=None)
-
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, Py_None) < 0) __PYX_ERR(0, 558, __pyx_L1_error)
-
+0561:         tree.nodeSummary = currentNodeSummary
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 561, __pyx_L1_error)
-
 0562:
-
 0563:         # Divide sets for child nodes
-
+0564:         if tree.trueBranch or tree.falseBranch:
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 564, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 564, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (!__pyx_t_6) {
-  } else {
-    __pyx_t_5 = __pyx_t_6;
-    goto __pyx_L4_bool_binop_done;
-  }
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 564, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_6 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_6 < 0)) __PYX_ERR(0, 564, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_5 = __pyx_t_6;
-  __pyx_L4_bool_binop_done:;
-  if (__pyx_t_5) {
-/* … */
-  }
-
+0565:             X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, tree.col, tree.value)
-
    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 565, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 565, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 565, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_7 = NULL;
-    __pyx_t_8 = 0;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
-      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_3);
-      if (likely(__pyx_t_7)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-        __Pyx_INCREF(__pyx_t_7);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_3, function);
-        __pyx_t_8 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_3)) {
-      PyObject *__pyx_temp[6] = {__pyx_t_7, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
-      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 5+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
-      PyObject *__pyx_temp[6] = {__pyx_t_7, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_t_2, __pyx_t_1};
-      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_8, 5+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    } else
-    #endif
-    {
-      __pyx_t_9 = PyTuple_New(5+__pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 565, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      if (__pyx_t_7) {
-        __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_t_7); __pyx_t_7 = NULL;
-      }
-      __Pyx_INCREF(__pyx_v_X);
-      __Pyx_GIVEREF(__pyx_v_X);
-      PyTuple_SET_ITEM(__pyx_t_9, 0+__pyx_t_8, __pyx_v_X);
-      __Pyx_INCREF(__pyx_v_treatment_idx);
-      __Pyx_GIVEREF(__pyx_v_treatment_idx);
-      PyTuple_SET_ITEM(__pyx_t_9, 1+__pyx_t_8, __pyx_v_treatment_idx);
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_9, 2+__pyx_t_8, __pyx_v_y);
-      __Pyx_GIVEREF(__pyx_t_2);
-      PyTuple_SET_ITEM(__pyx_t_9, 3+__pyx_t_8, __pyx_t_2);
-      __Pyx_GIVEREF(__pyx_t_1);
-      PyTuple_SET_ITEM(__pyx_t_9, 4+__pyx_t_8, __pyx_t_1);
-      __pyx_t_2 = 0;
-      __pyx_t_1 = 0;
-      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_9, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 565, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
-      PyObject* sequence = __pyx_t_4;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 6)) {
-        if (size > 6) __Pyx_RaiseTooManyValuesError(6);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 565, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_9 = PyTuple_GET_ITEM(sequence, 1); 
-        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 2); 
-        __pyx_t_2 = PyTuple_GET_ITEM(sequence, 3); 
-        __pyx_t_7 = PyTuple_GET_ITEM(sequence, 4); 
-        __pyx_t_10 = PyTuple_GET_ITEM(sequence, 5); 
-      } else {
-        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_9 = PyList_GET_ITEM(sequence, 1); 
-        __pyx_t_1 = PyList_GET_ITEM(sequence, 2); 
-        __pyx_t_2 = PyList_GET_ITEM(sequence, 3); 
-        __pyx_t_7 = PyList_GET_ITEM(sequence, 4); 
-        __pyx_t_10 = PyList_GET_ITEM(sequence, 5); 
-      }
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_9);
-      __Pyx_INCREF(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_7);
-      __Pyx_INCREF(__pyx_t_10);
-      #else
-      {
-        Py_ssize_t i;
-        PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_9,&__pyx_t_1,&__pyx_t_2,&__pyx_t_7,&__pyx_t_10};
-        for (i=0; i < 6; i++) {
-          PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 565, __pyx_L1_error)
-          __Pyx_GOTREF(item);
-          *(temps[i]) = item;
-        }
-      }
-      #endif
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      PyObject** temps[6] = {&__pyx_t_3,&__pyx_t_9,&__pyx_t_1,&__pyx_t_2,&__pyx_t_7,&__pyx_t_10};
-      __pyx_t_11 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 565, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_12 = Py_TYPE(__pyx_t_11)->tp_iternext;
-      for (index=0; index < 6; index++) {
-        PyObject* item = __pyx_t_12(__pyx_t_11); if (unlikely(!item)) goto __pyx_L6_unpacking_failed;
-        __Pyx_GOTREF(item);
-        *(temps[index]) = item;
-      }
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_11), 6) < 0) __PYX_ERR(0, 565, __pyx_L1_error)
-      __pyx_t_12 = NULL;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      goto __pyx_L7_unpacking_done;
-      __pyx_L6_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __pyx_t_12 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 565, __pyx_L1_error)
-      __pyx_L7_unpacking_done:;
-    }
-    __pyx_v_X_l = __pyx_t_3;
-    __pyx_t_3 = 0;
-    __pyx_v_X_r = __pyx_t_9;
-    __pyx_t_9 = 0;
-    __pyx_v_w_l = __pyx_t_1;
-    __pyx_t_1 = 0;
-    __pyx_v_w_r = __pyx_t_2;
-    __pyx_t_2 = 0;
-    __pyx_v_y_l = __pyx_t_7;
-    __pyx_t_7 = 0;
-    __pyx_v_y_r = __pyx_t_10;
-    __pyx_t_10 = 0;
-
 0566:
-
 0567:             # recursive call for each branch
-
+0568:             if tree.trueBranch is not None:
-
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 568, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = (__pyx_t_4 != Py_None);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_6 = (__pyx_t_5 != 0);
-    if (__pyx_t_6) {
-/* … */
-    }
-
+0569:                 self.fillTree(X_l, w_l, y_l, tree.trueBranch)
-
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 569, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 569, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __pyx_t_2 = NULL;
-      __pyx_t_8 = 0;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-        __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_10);
-        if (likely(__pyx_t_2)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-          __Pyx_INCREF(__pyx_t_2);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_10, function);
-          __pyx_t_8 = 1;
-        }
-      }
-      #if CYTHON_FAST_PYCALL
-      if (PyFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[5] = {__pyx_t_2, __pyx_v_X_l, __pyx_v_w_l, __pyx_v_y_l, __pyx_t_7};
-        __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      } else
-      #endif
-      #if CYTHON_FAST_PYCCALL
-      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[5] = {__pyx_t_2, __pyx_v_X_l, __pyx_v_w_l, __pyx_v_y_l, __pyx_t_7};
-        __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      } else
-      #endif
-      {
-        __pyx_t_1 = PyTuple_New(4+__pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 569, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        if (__pyx_t_2) {
-          __Pyx_GIVEREF(__pyx_t_2); PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_t_2); __pyx_t_2 = NULL;
-        }
-        __Pyx_INCREF(__pyx_v_X_l);
-        __Pyx_GIVEREF(__pyx_v_X_l);
-        PyTuple_SET_ITEM(__pyx_t_1, 0+__pyx_t_8, __pyx_v_X_l);
-        __Pyx_INCREF(__pyx_v_w_l);
-        __Pyx_GIVEREF(__pyx_v_w_l);
-        PyTuple_SET_ITEM(__pyx_t_1, 1+__pyx_t_8, __pyx_v_w_l);
-        __Pyx_INCREF(__pyx_v_y_l);
-        __Pyx_GIVEREF(__pyx_v_y_l);
-        PyTuple_SET_ITEM(__pyx_t_1, 2+__pyx_t_8, __pyx_v_y_l);
-        __Pyx_GIVEREF(__pyx_t_7);
-        PyTuple_SET_ITEM(__pyx_t_1, 3+__pyx_t_8, __pyx_t_7);
-        __pyx_t_7 = 0;
-        __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_1, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 569, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0570:             if tree.falseBranch is not None:
-
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 570, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_6 = (__pyx_t_4 != Py_None);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_5 = (__pyx_t_6 != 0);
-    if (__pyx_t_5) {
-/* … */
-    }
-
+0571:                 self.fillTree(X_r, w_r, y_r, tree.falseBranch)
-
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fillTree); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 571, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 571, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_7 = NULL;
-      __pyx_t_8 = 0;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-        __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_10);
-        if (likely(__pyx_t_7)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-          __Pyx_INCREF(__pyx_t_7);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_10, function);
-          __pyx_t_8 = 1;
-        }
-      }
-      #if CYTHON_FAST_PYCALL
-      if (PyFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[5] = {__pyx_t_7, __pyx_v_X_r, __pyx_v_w_r, __pyx_v_y_r, __pyx_t_1};
-        __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      } else
-      #endif
-      #if CYTHON_FAST_PYCCALL
-      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[5] = {__pyx_t_7, __pyx_v_X_r, __pyx_v_w_r, __pyx_v_y_r, __pyx_t_1};
-        __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 4+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      } else
-      #endif
-      {
-        __pyx_t_2 = PyTuple_New(4+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 571, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        if (__pyx_t_7) {
-          __Pyx_GIVEREF(__pyx_t_7); PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_7); __pyx_t_7 = NULL;
-        }
-        __Pyx_INCREF(__pyx_v_X_r);
-        __Pyx_GIVEREF(__pyx_v_X_r);
-        PyTuple_SET_ITEM(__pyx_t_2, 0+__pyx_t_8, __pyx_v_X_r);
-        __Pyx_INCREF(__pyx_v_w_r);
-        __Pyx_GIVEREF(__pyx_v_w_r);
-        PyTuple_SET_ITEM(__pyx_t_2, 1+__pyx_t_8, __pyx_v_w_r);
-        __Pyx_INCREF(__pyx_v_y_r);
-        __Pyx_GIVEREF(__pyx_v_y_r);
-        PyTuple_SET_ITEM(__pyx_t_2, 2+__pyx_t_8, __pyx_v_y_r);
-        __Pyx_GIVEREF(__pyx_t_1);
-        PyTuple_SET_ITEM(__pyx_t_2, 3+__pyx_t_8, __pyx_t_1);
-        __pyx_t_1 = 0;
-        __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_2, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 571, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
 0572:
-
 0573:         # Update Information
-
 0574:
-
 0575:         # matchScore
-
+0576:         matchScore = (currentNodeSummary[tree.bestTreatment][0] - currentNodeSummary[0][0])
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_bestTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_10 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_t_4); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = PyNumber_Subtract(__pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 576, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_matchScore = __pyx_t_10;
-  __pyx_t_10 = 0;
-
+0577:         tree.matchScore = round(matchScore, 4)
-
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 577, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_INCREF(__pyx_v_matchScore);
-  __Pyx_GIVEREF(__pyx_v_matchScore);
-  PyTuple_SET_ITEM(__pyx_t_10, 0, __pyx_v_matchScore);
-  __Pyx_INCREF(__pyx_int_4);
-  __Pyx_GIVEREF(__pyx_int_4);
-  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
-  __pyx_t_2 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 577, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_matchScore, __pyx_t_2) < 0) __PYX_ERR(0, 577, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0578:         tree.summary['matchScore'] = round(matchScore, 4)
-
  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 578, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_matchScore);
-  __Pyx_GIVEREF(__pyx_v_matchScore);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_matchScore);
-  __Pyx_INCREF(__pyx_int_4);
-  __Pyx_GIVEREF(__pyx_int_4);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_int_4);
-  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_2, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 578, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 578, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_n_s_matchScore, __pyx_t_10) < 0)) __PYX_ERR(0, 578, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
 0579:
-
 0580:         # Samples, Group_size
-
+0581:         tree.summary['samples'] = len(y)
-
  __pyx_t_13 = PyObject_Length(__pyx_v_y); if (unlikely(__pyx_t_13 == ((Py_ssize_t)-1))) __PYX_ERR(0, 581, __pyx_L1_error)
-  __pyx_t_10 = PyInt_FromSsize_t(__pyx_t_13); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 581, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 581, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (unlikely(PyObject_SetItem(__pyx_t_2, __pyx_n_s_samples, __pyx_t_10) < 0)) __PYX_ERR(0, 581, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
+0582:         tree.summary['group_size'] = ''
-
  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 582, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  if (unlikely(PyObject_SetItem(__pyx_t_10, __pyx_n_s_group_size, __pyx_kp_s__4) < 0)) __PYX_ERR(0, 582, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
+0583:         for treatment_group, summary in zip(self.classes_, currentNodeSummary):
-
  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __pyx_t_2 = PyTuple_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 583, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GIVEREF(__pyx_t_10);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_t_10);
-  __Pyx_INCREF(__pyx_v_currentNodeSummary);
-  __Pyx_GIVEREF(__pyx_v_currentNodeSummary);
-  PyTuple_SET_ITEM(__pyx_t_2, 1, __pyx_v_currentNodeSummary);
-  __pyx_t_10 = 0;
-  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_builtin_zip, __pyx_t_2, NULL); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_10)) || PyTuple_CheckExact(__pyx_t_10)) {
-    __pyx_t_2 = __pyx_t_10; __Pyx_INCREF(__pyx_t_2); __pyx_t_13 = 0;
-    __pyx_t_14 = NULL;
-  } else {
-    __pyx_t_13 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_10); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 583, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_14 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_14)) __PYX_ERR(0, 583, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_14)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_13 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_10 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_13); __Pyx_INCREF(__pyx_t_10); __pyx_t_13++; if (unlikely(0 < 0)) __PYX_ERR(0, 583, __pyx_L1_error)
-        #else
-        __pyx_t_10 = PySequence_ITEM(__pyx_t_2, __pyx_t_13); __pyx_t_13++; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        #endif
-      } else {
-        if (__pyx_t_13 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_10 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_13); __Pyx_INCREF(__pyx_t_10); __pyx_t_13++; if (unlikely(0 < 0)) __PYX_ERR(0, 583, __pyx_L1_error)
-        #else
-        __pyx_t_10 = PySequence_ITEM(__pyx_t_2, __pyx_t_13); __pyx_t_13++; if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 583, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_10);
-        #endif
-      }
-    } else {
-      __pyx_t_10 = __pyx_t_14(__pyx_t_2);
-      if (unlikely(!__pyx_t_10)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 583, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_10);
-    }
-    if ((likely(PyTuple_CheckExact(__pyx_t_10))) || (PyList_CheckExact(__pyx_t_10))) {
-      PyObject* sequence = __pyx_t_10;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 2)) {
-        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 583, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_4 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_1 = PyTuple_GET_ITEM(sequence, 1); 
-      } else {
-        __pyx_t_4 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_1 = PyList_GET_ITEM(sequence, 1); 
-      }
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(__pyx_t_1);
-      #else
-      __pyx_t_4 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 583, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_1 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 583, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      #endif
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      __pyx_t_7 = PyObject_GetIter(__pyx_t_10); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 583, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      __pyx_t_12 = Py_TYPE(__pyx_t_7)->tp_iternext;
-      index = 0; __pyx_t_4 = __pyx_t_12(__pyx_t_7); if (unlikely(!__pyx_t_4)) goto __pyx_L12_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_4);
-      index = 1; __pyx_t_1 = __pyx_t_12(__pyx_t_7); if (unlikely(!__pyx_t_1)) goto __pyx_L12_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_1);
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_12(__pyx_t_7), 2) < 0) __PYX_ERR(0, 583, __pyx_L1_error)
-      __pyx_t_12 = NULL;
-      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      goto __pyx_L13_unpacking_done;
-      __pyx_L12_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      __pyx_t_12 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 583, __pyx_L1_error)
-      __pyx_L13_unpacking_done:;
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_4);
-    __pyx_t_4 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_summary, __pyx_t_1);
-    __pyx_t_1 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0584:             tree.summary['group_size'] += ' ' + treatment_group + ': ' + str(summary[1])
-
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_summary); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __Pyx_INCREF(__pyx_n_s_group_size);
-    __pyx_t_15 = __pyx_n_s_group_size;
-    __pyx_t_1 = __Pyx_PyObject_Dict_GetItem(__pyx_t_10, __pyx_t_15); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = PyNumber_Add(__pyx_kp_s__5, __pyx_v_treatment_group); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_7 = PyNumber_Add(__pyx_t_4, __pyx_kp_s__6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = __Pyx_GetItemInt(__pyx_v_summary, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyString_Type)), __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = PyNumber_Add(__pyx_t_7, __pyx_t_9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_1, __pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (unlikely(PyObject_SetItem(__pyx_t_10, __pyx_t_15, __pyx_t_9) < 0)) __PYX_ERR(0, 584, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __Pyx_DECREF(__pyx_t_15); __pyx_t_15 = 0;
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
 0585:         # classProb
-
+0586:         if tree.results is not None:
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 586, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = (__pyx_t_2 != Py_None);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_6 = (__pyx_t_5 != 0);
-  if (__pyx_t_6) {
-/* … */
-  }
-
+0587:             tree.results = self.uplift_classification_results(treatment_idx, y)
-
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 587, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_9 = NULL;
-    __pyx_t_8 = 0;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-      __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_10);
-      if (likely(__pyx_t_9)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-        __Pyx_INCREF(__pyx_t_9);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_10, function);
-        __pyx_t_8 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_10)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_v_treatment_idx, __pyx_v_y};
-      __pyx_t_2 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __Pyx_GOTREF(__pyx_t_2);
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_9, __pyx_v_treatment_idx, __pyx_v_y};
-      __pyx_t_2 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_8, 2+__pyx_t_8); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __Pyx_GOTREF(__pyx_t_2);
-    } else
-    #endif
-    {
-      __pyx_t_4 = PyTuple_New(2+__pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 587, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      if (__pyx_t_9) {
-        __Pyx_GIVEREF(__pyx_t_9); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_9); __pyx_t_9 = NULL;
-      }
-      __Pyx_INCREF(__pyx_v_treatment_idx);
-      __Pyx_GIVEREF(__pyx_v_treatment_idx);
-      PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_8, __pyx_v_treatment_idx);
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_8, __pyx_v_y);
-      __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_4, NULL); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 587, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_tree, __pyx_n_s_results, __pyx_t_2) < 0) __PYX_ERR(0, 587, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0588:         return self
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_self);
-  __pyx_r = __pyx_v_self;
-  goto __pyx_L0;
-
 0589:
-
+0590:     def predict(self, X):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict[] = "\n        Returns the recommended treatment group and predicted optimal\n        probability conditional on using the recommended treatment group.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        Returns\n        -------\n        pred: ndarray, shape = [num_samples, num_treatments]\n            An ndarray of predicted treatment effects across treatments.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,0};
-    PyObject* values[2] = {0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, 1); __PYX_ERR(0, 590, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 590, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("predict", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 590, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict(__pyx_self, __pyx_v_self, __pyx_v_X);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_12predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X) {
-  PyObject *__pyx_v_pred_nodes = NULL;
-  Py_ssize_t __pyx_v_i_row;
-  PyObject *__pyx_v_pred_leaf = NULL;
-  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict", 0);
-  __Pyx_INCREF(__pyx_v_X);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_pred_nodes);
-  __Pyx_XDECREF(__pyx_v_pred_leaf);
-  __Pyx_XDECREF(__pyx_v__);
-  __Pyx_XDECREF(__pyx_v_X);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__35 = PyTuple_Pack(6, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_pred_nodes, __pyx_n_s_i_row, __pyx_n_s_pred_leaf, __pyx_n_s__34); if (unlikely(!__pyx_tuple__35)) __PYX_ERR(0, 590, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__35);
-  __Pyx_GIVEREF(__pyx_tuple__35);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_13predict, 0, __pyx_n_s_UpliftTreeClassifier_predict, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__36)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 590, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_predict, __pyx_t_2) < 0) __PYX_ERR(0, 590, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_codeobj__36 = (PyObject*)__Pyx_PyCode_New(2, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__35, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_predict, 590, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__36)) __PYX_ERR(0, 590, __pyx_L1_error)
-
 0591:         '''
-
 0592:         Returns the recommended treatment group and predicted optimal
-
 0593:         probability conditional on using the recommended treatment group.
-
 0594:
-
 0595:         Args
-
 0596:         ----
-
 0597:         X : ndarray, shape = [num_samples, num_features]
-
 0598:             An ndarray of the covariates used to train the uplift model.
-
 0599:
-
 0600:         Returns
-
 0601:         -------
-
 0602:         pred: ndarray, shape = [num_samples, num_treatments]
-
 0603:             An ndarray of predicted treatment effects across treatments.
-
 0604:         '''
-
 0605:
-
+0606:         X = check_array(X)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_array); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 606, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_3, __pyx_v_X) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_X);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 606, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF_SET(__pyx_v_X, __pyx_t_1);
-  __pyx_t_1 = 0;
-
 0607:
-
+0608:         pred_nodes = []
-
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 608, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_pred_nodes = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
+0609:         for i_row in range(len(X)):
-
  __pyx_t_4 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_4 == ((Py_ssize_t)-1))) __PYX_ERR(0, 609, __pyx_L1_error)
-  __pyx_t_5 = __pyx_t_4;
-  for (__pyx_t_6 = 0; __pyx_t_6 < __pyx_t_5; __pyx_t_6+=1) {
-    __pyx_v_i_row = __pyx_t_6;
-
+0610:             pred_leaf, _ = self.classify(X[i_row], self.fitted_uplift_tree, dataMissing=False)
-
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classify); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_X, __pyx_v_i_row, Py_ssize_t, 1, PyInt_FromSsize_t, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_fitted_uplift_tree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_7 = PyTuple_New(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __Pyx_GIVEREF(__pyx_t_2);
-    PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_2);
-    __Pyx_GIVEREF(__pyx_t_3);
-    PyTuple_SET_ITEM(__pyx_t_7, 1, __pyx_t_3);
-    __pyx_t_2 = 0;
-    __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_dataMissing, Py_False) < 0) __PYX_ERR(0, 610, __pyx_L1_error)
-    __pyx_t_2 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_7, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 610, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if ((likely(PyTuple_CheckExact(__pyx_t_2))) || (PyList_CheckExact(__pyx_t_2))) {
-      PyObject* sequence = __pyx_t_2;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 2)) {
-        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 610, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_7 = PyTuple_GET_ITEM(sequence, 1); 
-      } else {
-        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_7 = PyList_GET_ITEM(sequence, 1); 
-      }
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_7);
-      #else
-      __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 610, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_7 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 610, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      #endif
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 610, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __pyx_t_8 = Py_TYPE(__pyx_t_1)->tp_iternext;
-      index = 0; __pyx_t_3 = __pyx_t_8(__pyx_t_1); if (unlikely(!__pyx_t_3)) goto __pyx_L5_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_3);
-      index = 1; __pyx_t_7 = __pyx_t_8(__pyx_t_1); if (unlikely(!__pyx_t_7)) goto __pyx_L5_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_7);
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_8(__pyx_t_1), 2) < 0) __PYX_ERR(0, 610, __pyx_L1_error)
-      __pyx_t_8 = NULL;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      goto __pyx_L6_unpacking_done;
-      __pyx_L5_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_t_8 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 610, __pyx_L1_error)
-      __pyx_L6_unpacking_done:;
-    }
-    __Pyx_XDECREF_SET(__pyx_v_pred_leaf, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_7);
-    __pyx_t_7 = 0;
-
+0611:             pred_nodes.append(pred_leaf)
-
    __pyx_t_9 = __Pyx_PyList_Append(__pyx_v_pred_nodes, __pyx_v_pred_leaf); if (unlikely(__pyx_t_9 == ((int)-1))) __PYX_ERR(0, 611, __pyx_L1_error)
-  }
-
+0612:         return np.array(pred_nodes)
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_np); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 612, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_array); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 612, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_t_7 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_7)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_7);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_7, __pyx_v_pred_nodes) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_pred_nodes);
-  __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 612, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_r = __pyx_t_2;
-  __pyx_t_2 = 0;
-  goto __pyx_L0;
-
 0613:
-
+0614:     @staticmethod
-
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 614, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_divideSet, __pyx_t_3) < 0) __PYX_ERR(0, 615, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0615:     def divideSet(X, treatment_idx, y, column, value):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet[] = "\n        Tree node split.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        column : int\n                The column used to split the data.\n        value : float or int\n                The value in the column for splitting the data.\n\n        Returns\n        -------\n        (X_l, X_r, treatment_l, treatment_r, y_l, y_r) : list of ndarray\n                The covariates, treatments and outcomes of left node and the right node.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet = {"divideSet", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_column = 0;
-  PyObject *__pyx_v_value = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("divideSet (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_column,&__pyx_n_s_value,0};
-    PyObject* values[5] = {0,0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 1); __PYX_ERR(0, 615, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 2); __PYX_ERR(0, 615, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_column)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 3); __PYX_ERR(0, 615, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_value)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, 4); __PYX_ERR(0, 615, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "divideSet") < 0)) __PYX_ERR(0, 615, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 5) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-      values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-    }
-    __pyx_v_X = values[0];
-    __pyx_v_treatment_idx = values[1];
-    __pyx_v_y = values[2];
-    __pyx_v_column = values[3];
-    __pyx_v_value = values[4];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("divideSet", 1, 5, 5, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 615, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.divideSet", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet(__pyx_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_column, __pyx_v_value);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_14divideSet(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_column, PyObject *__pyx_v_value) {
-  PyObject *__pyx_v_filt = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("divideSet", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.divideSet", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_filt);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__37 = PyTuple_Pack(6, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_column, __pyx_n_s_value, __pyx_n_s_filt); if (unlikely(!__pyx_tuple__37)) __PYX_ERR(0, 615, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__37);
-  __Pyx_GIVEREF(__pyx_tuple__37);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_15divideSet, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_divideSet, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__38)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 615, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_codeobj__38 = (PyObject*)__Pyx_PyCode_New(5, 0, 6, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__37, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_divideSet, 615, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__38)) __PYX_ERR(0, 615, __pyx_L1_error)
-
 0616:         '''
-
 0617:         Tree node split.
-
 0618:
-
 0619:         Args
-
 0620:         ----
-
 0621:         X : ndarray, shape = [num_samples, num_features]
-
 0622:             An ndarray of the covariates used to train the uplift model.
-
 0623:         treatment_idx : array-like, shape = [num_samples]
-
 0624:             An array containing the treatment group index for each unit.
-
 0625:         y : array-like, shape = [num_samples]
-
 0626:             An array containing the outcome of interest for each unit.
-
 0627:         column : int
-
 0628:                 The column used to split the data.
-
 0629:         value : float or int
-
 0630:                 The value in the column for splitting the data.
-
 0631:
-
 0632:         Returns
-
 0633:         -------
-
 0634:         (X_l, X_r, treatment_l, treatment_r, y_l, y_r) : list of ndarray
-
 0635:                 The covariates, treatments and outcomes of left node and the right node.
-
 0636:         '''
-
 0637:         # for int and float values
-
+0638:         if np.issubdtype(value.dtype, np.number):
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 638, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_issubdtype); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 638, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_value, __pyx_n_s_dtype); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 638, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = NULL;
-  __pyx_t_5 = 0;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-      __pyx_t_5 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_3)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_2, ((PyObject *)__pyx_ptype_5numpy_number)};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_3)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_t_2, ((PyObject *)__pyx_ptype_5numpy_number)};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_3, __pyx_temp+1-__pyx_t_5, 2+__pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  } else
-  #endif
-  {
-    __pyx_t_6 = PyTuple_New(2+__pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 638, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    if (__pyx_t_4) {
-      __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_4); __pyx_t_4 = NULL;
-    }
-    __Pyx_GIVEREF(__pyx_t_2);
-    PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_5, __pyx_t_2);
-    __Pyx_INCREF(((PyObject *)__pyx_ptype_5numpy_number));
-    __Pyx_GIVEREF(((PyObject *)__pyx_ptype_5numpy_number));
-    PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_5, ((PyObject *)__pyx_ptype_5numpy_number));
-    __pyx_t_2 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_6, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 638, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_7 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_7 < 0)) __PYX_ERR(0, 638, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (__pyx_t_7) {
-/* … */
-    goto __pyx_L3;
-  }
-
+0639:             filt = X[:, column] >= value
-
    __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 639, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_INCREF(__pyx_slice__7);
-    __Pyx_GIVEREF(__pyx_slice__7);
-    PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_slice__7);
-    __Pyx_INCREF(__pyx_v_column);
-    __Pyx_GIVEREF(__pyx_v_column);
-    PyTuple_SET_ITEM(__pyx_t_1, 1, __pyx_v_column);
-    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 639, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = PyObject_RichCompare(__pyx_t_3, __pyx_v_value, Py_GE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 639, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_v_filt = __pyx_t_1;
-    __pyx_t_1 = 0;
-/* … */
-  __pyx_slice__7 = PySlice_New(Py_None, Py_None, Py_None); if (unlikely(!__pyx_slice__7)) __PYX_ERR(0, 639, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_slice__7);
-  __Pyx_GIVEREF(__pyx_slice__7);
-
 0640:         else:  # for strings
-
+0641:             filt = X[:, column] == value
-
  /*else*/ {
-    __pyx_t_1 = PyTuple_New(2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 641, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_INCREF(__pyx_slice__7);
-    __Pyx_GIVEREF(__pyx_slice__7);
-    PyTuple_SET_ITEM(__pyx_t_1, 0, __pyx_slice__7);
-    __Pyx_INCREF(__pyx_v_column);
-    __Pyx_GIVEREF(__pyx_v_column);
-    PyTuple_SET_ITEM(__pyx_t_1, 1, __pyx_v_column);
-    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 641, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = PyObject_RichCompare(__pyx_t_3, __pyx_v_value, Py_EQ); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 641, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_v_filt = __pyx_t_1;
-    __pyx_t_1 = 0;
-  }
-  __pyx_L3:;
-
 0642:
-
+0643:         return X[filt], X[~filt], treatment_idx[filt], treatment_idx[~filt], y[filt], y[~filt]
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_filt); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_6 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_treatment_idx, __pyx_v_filt); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_2 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_treatment_idx, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_filt); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_8 = PyNumber_Invert(__pyx_v_filt); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_9 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_t_8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  __pyx_t_8 = PyTuple_New(6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 643, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __Pyx_GIVEREF(__pyx_t_1);
-  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_t_1);
-  __Pyx_GIVEREF(__pyx_t_6);
-  PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_t_6);
-  __Pyx_GIVEREF(__pyx_t_3);
-  PyTuple_SET_ITEM(__pyx_t_8, 2, __pyx_t_3);
-  __Pyx_GIVEREF(__pyx_t_4);
-  PyTuple_SET_ITEM(__pyx_t_8, 3, __pyx_t_4);
-  __Pyx_GIVEREF(__pyx_t_2);
-  PyTuple_SET_ITEM(__pyx_t_8, 4, __pyx_t_2);
-  __Pyx_GIVEREF(__pyx_t_9);
-  PyTuple_SET_ITEM(__pyx_t_8, 5, __pyx_t_9);
-  __pyx_t_1 = 0;
-  __pyx_t_6 = 0;
-  __pyx_t_3 = 0;
-  __pyx_t_4 = 0;
-  __pyx_t_2 = 0;
-  __pyx_t_9 = 0;
-  __pyx_r = __pyx_t_8;
-  __pyx_t_8 = 0;
-  goto __pyx_L0;
-
 0644:
-
+0645:     def group_uniqueCounts(self, treatment_idx, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts[] = "\n        Count sample size by experiment group.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        results : list of list\n            The negative and positive outcome sample sizes for each of the control and treatment groups.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts = {"group_uniqueCounts", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("group_uniqueCounts (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,0};
-    PyObject* values[3] = {0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, 1); __PYX_ERR(0, 645, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, 2); __PYX_ERR(0, 645, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "group_uniqueCounts") < 0)) __PYX_ERR(0, 645, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_treatment_idx = values[1];
-    __pyx_v_y = values[2];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("group_uniqueCounts", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 645, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.group_uniqueCounts", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_16group_uniqueCounts(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_results = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_filt = NULL;
-  PyObject *__pyx_v_n_pos = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("group_uniqueCounts", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.group_uniqueCounts", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_results);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_filt);
-  __Pyx_XDECREF(__pyx_v_n_pos);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__39 = PyTuple_Pack(7, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_results, __pyx_n_s_i, __pyx_n_s_filt, __pyx_n_s_n_pos); if (unlikely(!__pyx_tuple__39)) __PYX_ERR(0, 645, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__39);
-  __Pyx_GIVEREF(__pyx_tuple__39);
-/* … */
-  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_17group_uniqueCounts, 0, __pyx_n_s_UpliftTreeClassifier_group_uniqu, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__40)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 645, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_group_uniqueCounts, __pyx_t_3) < 0) __PYX_ERR(0, 645, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_codeobj__40 = (PyObject*)__Pyx_PyCode_New(3, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__39, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_group_uniqueCounts, 645, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__40)) __PYX_ERR(0, 645, __pyx_L1_error)
-
 0646:         '''
-
 0647:         Count sample size by experiment group.
-
 0648:
-
 0649:         Args
-
 0650:         ----
-
 0651:         treatment_idx : array-like, shape = [num_samples]
-
 0652:             An array containing the treatment group index for each unit.
-
 0653:         y : array-like, shape = [num_samples]
-
 0654:             An array containing the outcome of interest for each unit.
-
 0655:
-
 0656:         Returns
-
 0657:         -------
-
 0658:         results : list of list
-
 0659:             The negative and positive outcome sample sizes for each of the control and treatment groups.
-
 0660:         '''
-
+0661:         results = []
-
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 661, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_results = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
+0662:         for i in range(self.n_class):
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 662, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 662, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 662, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 662, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 662, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 662, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 662, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0663:             filt = treatment_idx == i
-
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_treatment_idx, __pyx_v_i, Py_EQ); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 663, __pyx_L1_error)
-    __Pyx_XDECREF_SET(__pyx_v_filt, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0664:             n_pos = y[filt].sum()
-
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_filt); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 664, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_sum); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 664, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = NULL;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
-      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
-      if (likely(__pyx_t_5)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-        __Pyx_INCREF(__pyx_t_5);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_6, function);
-      }
-    }
-    __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_6);
-    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 664, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_2);
-    __pyx_t_2 = 0;
-
 0665:
-
 0666:             # [N(Y = 0, T = 1), N(Y = 1, T = 1)]
-
+0667:             results.append([filt.sum() - n_pos, n_pos])
-
    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_filt, __pyx_n_s_sum); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 667, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_5 = NULL;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
-      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
-      if (likely(__pyx_t_5)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-        __Pyx_INCREF(__pyx_t_5);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_6, function);
-      }
-    }
-    __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_6);
-    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 667, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_6 = PyNumber_Subtract(__pyx_t_2, __pyx_v_n_pos); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 667, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyList_New(2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 667, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_GIVEREF(__pyx_t_6);
-    PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_6);
-    __Pyx_INCREF(__pyx_v_n_pos);
-    __Pyx_GIVEREF(__pyx_v_n_pos);
-    PyList_SET_ITEM(__pyx_t_2, 1, __pyx_v_n_pos);
-    __pyx_t_6 = 0;
-    __pyx_t_7 = __Pyx_PyList_Append(__pyx_v_results, __pyx_t_2); if (unlikely(__pyx_t_7 == ((int)-1))) __PYX_ERR(0, 667, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 0668:
-
+0669:         return results
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_results);
-  __pyx_r = __pyx_v_results;
-  goto __pyx_L0;
-
 0670:
-
+0671:     @staticmethod
-
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 671, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_KL, __pyx_t_2) < 0) __PYX_ERR(0, 672, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0672:     def evaluate_KL(nodeSummary):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL[] = "\n        Calculate KL Divergence as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary()\n            method.\n\n        Returns\n        -------\n        d_res : KL Divergence\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL = {"evaluate_KL", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_KL (wrapper)", 0);
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_18evaluate_KL(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_v_p_c = NULL;
-  double __pyx_v_d_res;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_KL", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_KL", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_p_c);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__41 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_p_c, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__41)) __PYX_ERR(0, 672, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__41);
-  __Pyx_GIVEREF(__pyx_tuple__41);
-/* … */
-  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_19evaluate_KL, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_KL, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__42)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 672, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_codeobj__42 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__41, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_KL, 672, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__42)) __PYX_ERR(0, 672, __pyx_L1_error)
-
 0673:         '''
-
 0674:         Calculate KL Divergence as split evaluation criterion for a given node.
-
 0675:
-
 0676:         Args
-
 0677:         ----
-
 0678:         nodeSummary : list of list
-
 0679:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary()
-
 0680:             method.
-
 0681:
-
 0682:         Returns
-
 0683:         -------
-
 0684:         d_res : KL Divergence
-
 0685:         '''
-
+0686:         p_c = nodeSummary[0][0]
-
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 686, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 686, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_p_c = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+0687:         d_res = 0.
-
  __pyx_v_d_res = 0.;
-
+0688:         for treatment_group in nodeSummary[1:]:
-
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 688, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 688, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 688, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 688, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 688, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 688, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0689:             d_res += kl_divergence(treatment_group[0], p_c)
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 689, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_2); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_v_p_c); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
-    __pyx_t_7 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_t_5, __pyx_t_6); if (unlikely(__pyx_t_7 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 689, __pyx_L1_error)
-    __pyx_v_d_res = (__pyx_v_d_res + __pyx_t_7);
-
+0690:         return d_res
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = PyFloat_FromDouble(__pyx_v_d_res); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 690, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_r = __pyx_t_1;
-  __pyx_t_1 = 0;
-  goto __pyx_L0;
-
 0691:
-
+0692:     @staticmethod
-
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 692, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_ED, __pyx_t_3) < 0) __PYX_ERR(0, 693, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0693:     def evaluate_ED(nodeSummary):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED[] = "\n        Calculate Euclidean Distance as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : dictionary\n            The tree node summary statistics, produced by tree_node_summary()\n            method.\n\n        Returns\n        -------\n        d_res : Euclidean Distance\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED = {"evaluate_ED", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_ED (wrapper)", 0);
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_20evaluate_ED(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_v_pc = NULL;
-  PyObject *__pyx_v_d_res = NULL;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_ED", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_ED", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_pc);
-  __Pyx_XDECREF(__pyx_v_d_res);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__43 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__43)) __PYX_ERR(0, 693, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__43);
-  __Pyx_GIVEREF(__pyx_tuple__43);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_21evaluate_ED, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_ED, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__44)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 693, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_codeobj__44 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__43, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_ED, 693, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__44)) __PYX_ERR(0, 693, __pyx_L1_error)
-
 0694:         '''
-
 0695:         Calculate Euclidean Distance as split evaluation criterion for a given node.
-
 0696:
-
 0697:         Args
-
 0698:         ----
-
 0699:         nodeSummary : dictionary
-
 0700:             The tree node summary statistics, produced by tree_node_summary()
-
 0701:             method.
-
 0702:
-
 0703:         Returns
-
 0704:         -------
-
 0705:         d_res : Euclidean Distance
-
 0706:         '''
-
+0707:         pc = nodeSummary[0][0]
-
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 707, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 707, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_pc = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+0708:         d_res = 0
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_d_res = __pyx_int_0;
-
+0709:         for treatment_group in nodeSummary[1:]:
-
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 709, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 709, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 709, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 709, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 709, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 709, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0710:             d_res += 2*(treatment_group[0] - pc)**2
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 710, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyNumber_Power(__pyx_t_5, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = PyNumber_Multiply(__pyx_int_2, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 710, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 710, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0711:         return d_res
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_d_res);
-  __pyx_r = __pyx_v_d_res;
-  goto __pyx_L0;
-
 0712:
-
+0713:     @staticmethod
-
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 713, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_Chi, __pyx_t_2) < 0) __PYX_ERR(0, 714, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0714:     def evaluate_Chi(nodeSummary):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi[] = "\n        Calculate Chi-Square statistic as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : dictionary\n            The tree node summary statistics, produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Chi-Square\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi = {"evaluate_Chi", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_Chi (wrapper)", 0);
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_22evaluate_Chi(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_v_pc = NULL;
-  PyObject *__pyx_v_d_res = NULL;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_Chi", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_Chi", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_pc);
-  __Pyx_XDECREF(__pyx_v_d_res);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__45 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__45)) __PYX_ERR(0, 714, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__45);
-  __Pyx_GIVEREF(__pyx_tuple__45);
-/* … */
-  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_23evaluate_Chi, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_Ch, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__46)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 714, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_codeobj__46 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__45, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_Chi, 714, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__46)) __PYX_ERR(0, 714, __pyx_L1_error)
-
 0715:         '''
-
 0716:         Calculate Chi-Square statistic as split evaluation criterion for a given node.
-
 0717:
-
 0718:         Args
-
 0719:         ----
-
 0720:         nodeSummary : dictionary
-
 0721:             The tree node summary statistics, produced by tree_node_summary() method.
-
 0722:
-
 0723:         Returns
-
 0724:         -------
-
 0725:         d_res : Chi-Square
-
 0726:         '''
-
+0727:         pc = nodeSummary[0][0]
-
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 727, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 727, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_pc = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+0728:         d_res = 0
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_d_res = __pyx_int_0;
-
+0729:         for treatment_group in nodeSummary[1:]:
-
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 729, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 729, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 729, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 729, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 729, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 729, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0730:             d_res += ((treatment_group[0] - pc) ** 2 / max(0.1 ** 6, pc)
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyNumber_Power(__pyx_t_5, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_INCREF(__pyx_v_pc);
-    __pyx_t_5 = __pyx_v_pc;
-    __pyx_t_6 = pow(0.1, 6.0);
-    __pyx_t_8 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_9 = PyObject_RichCompare(__pyx_t_5, __pyx_t_8, Py_GT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    if (__pyx_t_10) {
-      __Pyx_INCREF(__pyx_t_5);
-      __pyx_t_7 = __pyx_t_5;
-    } else {
-      __pyx_t_9 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 730, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_7 = __pyx_t_9;
-      __pyx_t_9 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_7); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-/* … */
-    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_9); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 730, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0731:                       + (treatment_group[0] - pc) ** 2 / max(0.1 ** 6, 1 - pc))
-
    __pyx_t_7 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_2 = PyNumber_Subtract(__pyx_t_7, __pyx_v_pc); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __pyx_t_7 = PyNumber_Power(__pyx_t_2, __pyx_int_2, Py_None); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_pc, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_6 = pow(0.1, 6.0);
-    __pyx_t_8 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_11 = PyObject_RichCompare(__pyx_t_2, __pyx_t_8, Py_GT); __Pyx_XGOTREF(__pyx_t_11); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    __pyx_t_10 = __Pyx_PyObject_IsTrue(__pyx_t_11); if (unlikely(__pyx_t_10 < 0)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-    if (__pyx_t_10) {
-      __Pyx_INCREF(__pyx_t_2);
-      __pyx_t_9 = __pyx_t_2;
-    } else {
-      __pyx_t_11 = PyFloat_FromDouble(__pyx_t_6); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 731, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_9 = __pyx_t_11;
-      __pyx_t_11 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_7, __pyx_t_9); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_t_9 = PyNumber_Add(__pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 731, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0732:         return d_res
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_d_res);
-  __pyx_r = __pyx_v_d_res;
-  goto __pyx_L0;
-
 0733:
-
+0734:     @staticmethod
-
  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 734, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_DDP, __pyx_t_3) < 0) __PYX_ERR(0, 735, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+0735:     def evaluate_DDP(nodeSummary):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP[] = "\n        Calculate Delta P as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Delta P\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP = {"evaluate_DDP", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_DDP (wrapper)", 0);
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_24evaluate_DDP(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_v_pc = NULL;
-  PyObject *__pyx_v_d_res = NULL;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_DDP", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_DDP", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_pc);
-  __Pyx_XDECREF(__pyx_v_d_res);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__47 = PyTuple_Pack(4, __pyx_n_s_nodeSummary, __pyx_n_s_pc, __pyx_n_s_d_res, __pyx_n_s_treatment_group); if (unlikely(!__pyx_tuple__47)) __PYX_ERR(0, 735, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__47);
-  __Pyx_GIVEREF(__pyx_tuple__47);
-/* … */
-  __pyx_t_2 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_25evaluate_DDP, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_DD, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__48)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 735, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_codeobj__48 = (PyObject*)__Pyx_PyCode_New(1, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__47, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_DDP, 735, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__48)) __PYX_ERR(0, 735, __pyx_L1_error)
-
 0736:         '''
-
 0737:         Calculate Delta P as split evaluation criterion for a given node.
-
 0738:
-
 0739:         Args
-
 0740:         ----
-
 0741:         nodeSummary : list of list
-
 0742:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.
-
 0743:
-
 0744:         Returns
-
 0745:         -------
-
 0746:         d_res : Delta P
-
 0747:         '''
-
+0748:         pc = nodeSummary[0][0]
-
  __pyx_t_1 = __Pyx_GetItemInt(__pyx_v_nodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 748, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_GetItemInt(__pyx_t_1, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 748, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_pc = __pyx_t_2;
-  __pyx_t_2 = 0;
-
+0749:         d_res = 0
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_d_res = __pyx_int_0;
-
+0750:         for treatment_group in nodeSummary[1:]:
-
  __pyx_t_2 = __Pyx_PyObject_GetSlice(__pyx_v_nodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-    __pyx_t_1 = __pyx_t_2; __Pyx_INCREF(__pyx_t_1); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 750, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 750, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 750, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_3); __Pyx_INCREF(__pyx_t_2); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 750, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 750, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_4(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 750, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0751:             d_res += treatment_group[0] - pc
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_treatment_group, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 751, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PyNumber_Subtract(__pyx_t_2, __pyx_v_pc); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 751, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyNumber_InPlaceAdd(__pyx_v_d_res, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 751, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_DECREF_SET(__pyx_v_d_res, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0752:         return d_res
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_d_res);
-  __pyx_r = __pyx_v_d_res;
-  goto __pyx_L0;
-
 0753:
-
+0754:     @staticmethod
-
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 754, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_evaluate_CTS, __pyx_t_2) < 0) __PYX_ERR(0, 755, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+0755:     def evaluate_CTS(nodeSummary):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS[] = "\n        Calculate CTS (conditional treatment selection) as split evaluation criterion for a given node.\n\n        Args\n        ----\n        nodeSummary : list of list\n            The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.\n\n        Returns\n        -------\n        d_res : Chi-Square\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS = {"evaluate_CTS", (PyCFunction)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS, METH_O, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS(PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_CTS (wrapper)", 0);
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS(__pyx_self, ((PyObject *)__pyx_v_nodeSummary));
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_26evaluate_CTS(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_nodeSummary) {
-  PyObject *__pyx_v_stat = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("evaluate_CTS", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.evaluate_CTS", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_stat);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__49 = PyTuple_Pack(2, __pyx_n_s_nodeSummary, __pyx_n_s_stat); if (unlikely(!__pyx_tuple__49)) __PYX_ERR(0, 755, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__49);
-  __Pyx_GIVEREF(__pyx_tuple__49);
-/* … */
-  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_27evaluate_CTS, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_evaluate_CT, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__50)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 755, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_codeobj__50 = (PyObject*)__Pyx_PyCode_New(1, 0, 2, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__49, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_evaluate_CTS, 755, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__50)) __PYX_ERR(0, 755, __pyx_L1_error)
-
 0756:         '''
-
 0757:         Calculate CTS (conditional treatment selection) as split evaluation criterion for a given node.
-
 0758:
-
 0759:         Args
-
 0760:         ----
-
 0761:         nodeSummary : list of list
-
 0762:             The tree node summary statistics, [P(Y=1|T), N(T)], produced by tree_node_summary() method.
-
 0763:
-
 0764:         Returns
-
 0765:         -------
-
 0766:         d_res : Chi-Square
-
 0767:         '''
-
+0768:         return -max([stat[0] for stat in nodeSummary])
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 768, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (likely(PyList_CheckExact(__pyx_v_nodeSummary)) || PyTuple_CheckExact(__pyx_v_nodeSummary)) {
-    __pyx_t_2 = __pyx_v_nodeSummary; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_nodeSummary); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 768, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 768, __pyx_L1_error)
-  }
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 768, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_5); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 768, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      }
-    } else {
-      __pyx_t_5 = __pyx_t_4(__pyx_t_2);
-      if (unlikely(!__pyx_t_5)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 768, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_5);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_stat, __pyx_t_5);
-    __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_stat, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 768, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_5))) __PYX_ERR(0, 768, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_max, __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 768, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = PyNumber_Negative(__pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 768, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_r = __pyx_t_1;
-  __pyx_t_1 = 0;
-  goto __pyx_L0;
-
 0769:
-
+0770:     def normI(self, n_c: cython.int, n_c_left: cython.int, n_t: list, n_t_left: list,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI[] = "\n        Normalization factor.\n\n        Args\n        ----\n        currentNodeSummary : list of list\n            The summary statistics of the current tree node, [P(Y=1|T), N(T)].\n\n        leftNodeSummary : list of list\n            The summary statistics of the left tree node, [P(Y=1|T), N(T)].\n\n        alpha : float\n            The weight used to balance different normalization parts.\n\n        Returns\n        -------\n        norm_res : float\n            Normalization factor.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI = {"normI", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  CYTHON_UNUSED PyObject *__pyx_v_self = 0;
-  int __pyx_v_n_c;
-  int __pyx_v_n_c_left;
-  PyObject *__pyx_v_n_t = 0;
-  PyObject *__pyx_v_n_t_left = 0;
-  float __pyx_v_alpha;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("normI (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_n_c,&__pyx_n_s_n_c_left,&__pyx_n_s_n_t,&__pyx_n_s_n_t_left,&__pyx_n_s_alpha,0};
-    PyObject* values[6] = {0,0,0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_c)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 1); __PYX_ERR(0, 770, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_c_left)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 2); __PYX_ERR(0, 770, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_t)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 3); __PYX_ERR(0, 770, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (likely((values[4] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_t_left)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, 4); __PYX_ERR(0, 770, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_alpha);
-          if (value) { values[5] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "normI") < 0)) __PYX_ERR(0, 770, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_n_c = __Pyx_PyInt_As_int(values[1]); if (unlikely((__pyx_v_n_c == (int)-1) && PyErr_Occurred())) __PYX_ERR(0, 770, __pyx_L3_error)
-    __pyx_v_n_c_left = __Pyx_PyInt_As_int(values[2]); if (unlikely((__pyx_v_n_c_left == (int)-1) && PyErr_Occurred())) __PYX_ERR(0, 770, __pyx_L3_error)
-    __pyx_v_n_t = ((PyObject*)values[3]);
-    __pyx_v_n_t_left = ((PyObject*)values[4]);
-    if (values[5]) {
-      __pyx_v_alpha = __pyx_PyFloat_AsFloat(values[5]); if (unlikely((__pyx_v_alpha == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 771, __pyx_L3_error)
-    } else {
-      __pyx_v_alpha = ((float)((double)0.9));
-    }
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("normI", 0, 5, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 770, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.normI", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_n_t), (&PyList_Type), 1, "n_t", 1))) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_n_t_left), (&PyList_Type), 1, "n_t_left", 1))) __PYX_ERR(0, 770, __pyx_L1_error)
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI(__pyx_self, __pyx_v_self, __pyx_v_n_c, __pyx_v_n_c_left, __pyx_v_n_t, __pyx_v_n_t_left, __pyx_v_alpha);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_28normI(CYTHON_UNUSED PyObject *__pyx_self, CYTHON_UNUSED PyObject *__pyx_v_self, int __pyx_v_n_c, int __pyx_v_n_c_left, PyObject *__pyx_v_n_t, PyObject *__pyx_v_n_t_left, float __pyx_v_alpha) {
-  float __pyx_v_norm_res;
-  float __pyx_v_pt_a;
-  float __pyx_v_pc_a;
-  Py_ssize_t __pyx_v_i;
-  PyObject *__pyx_v_pt_a_i = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("normI", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_12);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.normI", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_pt_a_i);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__51 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_n_c, __pyx_n_s_n_c_left, __pyx_n_s_n_t, __pyx_n_s_n_t_left, __pyx_n_s_alpha, __pyx_n_s_norm_res, __pyx_n_s_pt_a, __pyx_n_s_pc_a, __pyx_n_s_i, __pyx_n_s_pt_a_i); if (unlikely(!__pyx_tuple__51)) __PYX_ERR(0, 770, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__51);
-  __Pyx_GIVEREF(__pyx_tuple__51);
-/* … */
-  __pyx_t_3 = PyTuple_New(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 770, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_GIVEREF(__pyx_t_2);
-  PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_2);
-  __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 770, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_c, __pyx_n_u_int) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_c_left, __pyx_n_u_int) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_t, __pyx_n_u_list) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_n_t_left, __pyx_n_u_list) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_alpha, __pyx_n_u_float) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_return, __pyx_n_u_float) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_29normI, 0, __pyx_n_s_UpliftTreeClassifier_normI, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__52)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 770, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_t_3);
-  __Pyx_CyFunction_SetAnnotationsDict(__pyx_t_5, __pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_normI, __pyx_t_5) < 0) __PYX_ERR(0, 770, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_codeobj__52 = (PyObject*)__Pyx_PyCode_New(6, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__51, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_normI, 770, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__52)) __PYX_ERR(0, 770, __pyx_L1_error)
-
+0771:               alpha: cython.float=0.9) -> cython.float:
-
  __pyx_t_2 = PyFloat_FromDouble(((double)0.9)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 771, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-
 0772:         '''
-
 0773:         Normalization factor.
-
 0774:
-
 0775:         Args
-
 0776:         ----
-
 0777:         currentNodeSummary : list of list
-
 0778:             The summary statistics of the current tree node, [P(Y=1|T), N(T)].
-
 0779:
-
 0780:         leftNodeSummary : list of list
-
 0781:             The summary statistics of the left tree node, [P(Y=1|T), N(T)].
-
 0782:
-
 0783:         alpha : float
-
 0784:             The weight used to balance different normalization parts.
-
 0785:
-
 0786:         Returns
-
 0787:         -------
-
 0788:         norm_res : float
-
 0789:             Normalization factor.
-
 0790:         '''
-
 0791:
-
+0792:         norm_res: cython.float = 0.
-
  __pyx_v_norm_res = 0.;
-
 0793:         pt_a: cython.float
-
 0794:         pc_a: cython.float
-
 0795:
-
+0796:         pt_a = 1. * np.sum(n_t_left) / (np.sum(n_t) + 0.1)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_2, __pyx_v_n_t_left) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t_left);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = PyNumber_Multiply(__pyx_float_1_, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_4))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_4);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_4, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_2) ? __Pyx_PyObject_Call2Args(__pyx_t_4, __pyx_t_2, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyFloat_AddObjC(__pyx_t_1, __pyx_float_0_1, 0.1, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 796, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_v_pt_a = __pyx_t_5;
-
+0797:         pc_a = 1. * n_c_left / (n_c + 0.1)
-
  __pyx_v_pc_a = ((1. * __pyx_v_n_c_left) / (__pyx_v_n_c + 0.1));
-
 0798:
-
 0799:         # Normalization Part 1
-
+0800:         norm_res += (
-
  __pyx_v_norm_res = (__pyx_v_norm_res + ((__pyx_v_alpha * __pyx_t_7) * __pyx_t_6));
-
+0801:             alpha * entropyH(1. * np.sum(n_t) / (np.sum(n_t) + n_c), 1. * n_c / (np.sum(n_t) + n_c))
-
  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = PyNumber_Multiply(__pyx_float_1_, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_4 = PyNumber_Add(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_2); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_sum); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_1);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_1, function);
-    }
-  }
-  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_3, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_3 = PyNumber_Add(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_8.__pyx_n = 1;
-  __pyx_t_8.q = __pyx_t_6;
-  __pyx_t_7 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_5, &__pyx_t_8); if (unlikely(__pyx_t_7 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 801, __pyx_L1_error)
-
+0802:             * kl_divergence(pt_a, pc_a)
-
  __pyx_t_6 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_v_pt_a, __pyx_v_pc_a); if (unlikely(__pyx_t_6 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 802, __pyx_L1_error)
-
 0803:         )
-
 0804:         # Normalization Part 2 & 3
-
+0805:         for i in range(len(n_t)):
-
  if (unlikely(__pyx_v_n_t == Py_None)) {
-    PyErr_SetString(PyExc_TypeError, "object of type 'NoneType' has no len()");
-    __PYX_ERR(0, 805, __pyx_L1_error)
-  }
-  __pyx_t_9 = PyList_GET_SIZE(__pyx_v_n_t); if (unlikely(__pyx_t_9 == ((Py_ssize_t)-1))) __PYX_ERR(0, 805, __pyx_L1_error)
-  __pyx_t_10 = __pyx_t_9;
-  for (__pyx_t_11 = 0; __pyx_t_11 < __pyx_t_10; __pyx_t_11+=1) {
-    __pyx_v_i = __pyx_t_11;
-
+0806:             pt_a_i = 1. * n_t_left[i] / (n_t[i] + 0.1)
-
    if (unlikely(__pyx_v_n_t_left == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 806, __pyx_L1_error)
-    }
-    __pyx_t_1 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t_left, __pyx_v_i)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 806, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    if (unlikely(__pyx_v_n_t == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 806, __pyx_L1_error)
-    }
-    __pyx_t_3 = __Pyx_PyFloat_AddObjC(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_float_0_1, 0.1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 806, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 806, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_pt_a_i, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0807:             norm_res += (
-
    __pyx_v_norm_res = (__pyx_v_norm_res + (((1.0 - __pyx_v_alpha) * __pyx_t_5) * __pyx_t_6));
-
+0808:                 (1 - alpha) * entropyH(1. * n_t[i] / (n_t[i] + n_c), 1. * n_c / (n_t[i] + n_c))
-
    if (unlikely(__pyx_v_n_t == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 808, __pyx_L1_error)
-    }
-    __pyx_t_2 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (unlikely(__pyx_v_n_t == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 808, __pyx_L1_error)
-    }
-    __pyx_t_3 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_1 = PyNumber_Add(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_t_3); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    if (unlikely(__pyx_v_n_t == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 808, __pyx_L1_error)
-    }
-    __pyx_t_1 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_2 = PyNumber_Add(PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i), __pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_1 = __Pyx_PyNumber_Divide(__pyx_t_3, __pyx_t_2); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_7 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_7 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_8.__pyx_n = 1;
-    __pyx_t_8.q = __pyx_t_7;
-    __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_6, &__pyx_t_8); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 808, __pyx_L1_error)
-
+0809:                 * kl_divergence(1. * pt_a_i, pc_a)
-
    __pyx_t_1 = PyNumber_Multiply(__pyx_float_1_, __pyx_v_pt_a_i); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 809, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_7 = __pyx_PyFloat_AsFloat(__pyx_t_1); if (unlikely((__pyx_t_7 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 809, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_t_6 = __pyx_f_8causalml_9inference_4tree_6uplift_kl_divergence(__pyx_t_7, __pyx_v_pc_a); if (unlikely(__pyx_t_6 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 809, __pyx_L1_error)
-
 0810:             )
-
+0811:             norm_res += (1. * n_t[i] / (np.sum(n_t) + n_c) * entropyH(pt_a_i))
-
    __pyx_t_1 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    if (unlikely(__pyx_v_n_t == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 811, __pyx_L1_error)
-    }
-    __pyx_t_2 = PyNumber_Multiply(__pyx_float_1_, PyList_GET_ITEM(__pyx_v_n_t, __pyx_v_i)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_GetModuleGlobalName(__pyx_t_4, __pyx_n_s_np); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_4, __pyx_n_s_sum); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_12))) {
-      __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_12);
-      if (likely(__pyx_t_4)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_12);
-        __Pyx_INCREF(__pyx_t_4);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_12, function);
-      }
-    }
-    __pyx_t_3 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_12, __pyx_t_4, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_12, __pyx_v_n_t);
-    __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __pyx_t_12 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __pyx_t_4 = PyNumber_Add(__pyx_t_3, __pyx_t_12); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __pyx_t_12 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_6 = __pyx_PyFloat_AsFloat(__pyx_v_pt_a_i); if (unlikely((__pyx_t_6 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
-    __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_t_6, NULL); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
-    __pyx_t_4 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_2 = PyNumber_Multiply(__pyx_t_12, __pyx_t_4); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = PyNumber_InPlaceAdd(__pyx_t_1, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_4); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 811, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_v_norm_res = __pyx_t_5;
-  }
-
 0812:         # Normalization Part 4
-
+0813:         norm_res += 1. * n_c/(np.sum(n_t) + n_c) * entropyH(pc_a)
-
  __pyx_t_4 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_2 = PyFloat_FromDouble((1. * __pyx_v_n_c)); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_np); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_12, __pyx_n_s_sum); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_t_12 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_12)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_12);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_12) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_12, __pyx_v_n_t) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyInt_From_int(__pyx_v_n_c); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_12 = PyNumber_Add(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyNumber_Divide(__pyx_t_2, __pyx_t_12); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_t_5 = __pyx_f_8causalml_9inference_4tree_6uplift_entropyH(__pyx_v_pc_a, NULL); if (unlikely(__pyx_t_5 == ((float)-1) && PyErr_Occurred())) __PYX_ERR(0, 813, __pyx_L1_error)
-  __pyx_t_12 = PyFloat_FromDouble(__pyx_t_5); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __pyx_t_2 = PyNumber_Multiply(__pyx_t_3, __pyx_t_12); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_t_12 = PyNumber_InPlaceAdd(__pyx_t_4, __pyx_t_2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_5 = __pyx_PyFloat_AsFloat(__pyx_t_12); if (unlikely((__pyx_t_5 == (float)-1) && PyErr_Occurred())) __PYX_ERR(0, 813, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_v_norm_res = __pyx_t_5;
-
 0814:
-
 0815:         # Normalization Part 5
-
+0816:         norm_res += 0.5
-
  __pyx_v_norm_res = (__pyx_v_norm_res + 0.5);
-
+0817:         return norm_res
-
  __Pyx_XDECREF(__pyx_r);
-  __pyx_t_12 = PyFloat_FromDouble(__pyx_v_norm_res); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 817, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __pyx_r = __pyx_t_12;
-  __pyx_t_12 = 0;
-  goto __pyx_L0;
-
 0818:
-
+0819:     def tree_node_summary(self, treatment_idx, y, min_samples_treatment=10, n_reg=100, parentNodeSummary=None):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary[] = "\n        Tree node summary statistics.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        min_samples_treatment: int, optional (default=10)\n            The minimum number of samples required of the experiment group t be split at a leaf node.\n        n_reg :  int, optional (default=10)\n            The regularization parameter defined in Rzepakowski et al. 2012,\n            the weight (in terms of sample size) of the parent node influence\n            on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : list of list\n            The positive probabilities and sample sizes of each of the control and treatment groups\n            in the parent node.\n\n        Returns\n        -------\n        nodeSummary : list of list\n            The positive probabilities and sample sizes of each of the control and treatment groups\n            in the current node.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary = {"tree_node_summary", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_min_samples_treatment = 0;
-  PyObject *__pyx_v_n_reg = 0;
-  PyObject *__pyx_v_parentNodeSummary = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("tree_node_summary (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
-    PyObject* values[6] = {0,0,0,0,0,0};
-    values[3] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[4] = ((PyObject *)((PyObject *)__pyx_int_100));
-    values[5] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, 1); __PYX_ERR(0, 819, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, 2); __PYX_ERR(0, 819, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
-          if (value) { values[3] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
-          if (value) { values[5] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "tree_node_summary") < 0)) __PYX_ERR(0, 819, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_treatment_idx = values[1];
-    __pyx_v_y = values[2];
-    __pyx_v_min_samples_treatment = values[3];
-    __pyx_v_n_reg = values[4];
-    __pyx_v_parentNodeSummary = values[5];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("tree_node_summary", 0, 3, 6, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 819, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.tree_node_summary", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_30tree_node_summary(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
-  PyObject *__pyx_v_counts = NULL;
-  PyObject *__pyx_v_nodeSummary = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_count = NULL;
-  PyObject *__pyx_v_n_pos = NULL;
-  PyObject *__pyx_v_n = NULL;
-  PyObject *__pyx_v_p = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("tree_node_summary", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.tree_node_summary", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_counts);
-  __Pyx_XDECREF(__pyx_v_nodeSummary);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_count);
-  __Pyx_XDECREF(__pyx_v_n_pos);
-  __Pyx_XDECREF(__pyx_v_n);
-  __Pyx_XDECREF(__pyx_v_p);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__53 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_counts, __pyx_n_s_nodeSummary, __pyx_n_s_i, __pyx_n_s_count, __pyx_n_s_n_pos, __pyx_n_s_n, __pyx_n_s_p); if (unlikely(!__pyx_tuple__53)) __PYX_ERR(0, 819, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__53);
-  __Pyx_GIVEREF(__pyx_tuple__53);
-  __pyx_codeobj__54 = (PyObject*)__Pyx_PyCode_New(6, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__53, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_tree_node_summary, 819, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__54)) __PYX_ERR(0, 819, __pyx_L1_error)
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_31tree_node_summary, 0, __pyx_n_s_UpliftTreeClassifier_tree_node_s, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__54)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 819, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__55);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_tree_node_summary, __pyx_t_5) < 0) __PYX_ERR(0, 819, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_tuple__55 = PyTuple_Pack(3, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__55)) __PYX_ERR(0, 819, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__55);
-  __Pyx_GIVEREF(__pyx_tuple__55);
-
 0820:         '''
-
 0821:         Tree node summary statistics.
-
 0822:
-
 0823:         Args
-
 0824:         ----
-
 0825:         treatment_idx : array-like, shape = [num_samples]
-
 0826:             An array containing the treatment group index for each unit.
-
 0827:         y : array-like, shape = [num_samples]
-
 0828:             An array containing the outcome of interest for each unit.
-
 0829:         min_samples_treatment: int, optional (default=10)
-
 0830:             The minimum number of samples required of the experiment group t be split at a leaf node.
-
 0831:         n_reg :  int, optional (default=10)
-
 0832:             The regularization parameter defined in Rzepakowski et al. 2012,
-
 0833:             the weight (in terms of sample size) of the parent node influence
-
 0834:             on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
-
 0835:         parentNodeSummary : list of list
-
 0836:             The positive probabilities and sample sizes of each of the control and treatment groups
-
 0837:             in the parent node.
-
 0838:
-
 0839:         Returns
-
 0840:         -------
-
 0841:         nodeSummary : list of list
-
 0842:             The positive probabilities and sample sizes of each of the control and treatment groups
-
 0843:             in the current node.
-
 0844:         '''
-
 0845:         # counts: [[N(Y=0, T=0), N(Y=1, T=0)], [N(Y=0, T=1), N(Y=1, T=1)], ...]
-
+0846:         counts = self.group_uniqueCounts(treatment_idx, y)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_group_uniqueCounts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 846, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 846, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_treatment_idx);
-    __Pyx_GIVEREF(__pyx_v_treatment_idx);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_treatment_idx);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 846, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_counts = __pyx_t_1;
-  __pyx_t_1 = 0;
-
 0847:
-
 0848:         # nodeSummary: [[P(Y=1|T=0), N(T=0)], [P(Y=1|T=1), N(T=1)], ...]
-
+0849:         nodeSummary = []
-
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 849, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_nodeSummary = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
 0850:         # Iterate the control and treatment groups
-
+0851:         for i, count in enumerate(counts):
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_t_1 = __pyx_int_0;
-  if (likely(PyList_CheckExact(__pyx_v_counts)) || PyTuple_CheckExact(__pyx_v_counts)) {
-    __pyx_t_2 = __pyx_v_counts; __Pyx_INCREF(__pyx_t_2); __pyx_t_6 = 0;
-    __pyx_t_7 = NULL;
-  } else {
-    __pyx_t_6 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_v_counts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 851, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_7 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 851, __pyx_L1_error)
-  }
-  for (;;) {
-    if (likely(!__pyx_t_7)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 851, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      } else {
-        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 851, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_2, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      }
-    } else {
-      __pyx_t_5 = __pyx_t_7(__pyx_t_2);
-      if (unlikely(!__pyx_t_5)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 851, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_5);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_count, __pyx_t_5);
-    __pyx_t_5 = 0;
-    __Pyx_INCREF(__pyx_t_1);
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_1);
-    __pyx_t_5 = __Pyx_PyInt_AddObjC(__pyx_t_1, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 851, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_1);
-    __pyx_t_1 = __pyx_t_5;
-    __pyx_t_5 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0852:             n_pos = count[1]
-
    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_count, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 852, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_5);
-    __pyx_t_5 = 0;
-
+0853:             n = count[0] + n_pos
-
    __pyx_t_5 = __Pyx_GetItemInt(__pyx_v_count, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 853, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_3 = PyNumber_Add(__pyx_t_5, __pyx_v_n_pos); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 853, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_n, __pyx_t_3);
-    __pyx_t_3 = 0;
-
+0854:             if parentNodeSummary is None:
-
    __pyx_t_8 = (__pyx_v_parentNodeSummary == Py_None);
-    __pyx_t_9 = (__pyx_t_8 != 0);
-    if (__pyx_t_9) {
-/* … */
-      goto __pyx_L5;
-    }
-
+0855:                 p = n_pos / n if n > 0 else 0.
-
      __pyx_t_5 = PyObject_RichCompare(__pyx_v_n, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 855, __pyx_L1_error)
-      __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 855, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (__pyx_t_9) {
-        __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_v_n_pos, __pyx_v_n); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 855, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __pyx_t_3 = __pyx_t_5;
-        __pyx_t_5 = 0;
-      } else {
-        __Pyx_INCREF(__pyx_float_0_);
-        __pyx_t_3 = __pyx_float_0_;
-      }
-      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_3);
-      __pyx_t_3 = 0;
-
+0856:             elif n > min_samples_treatment:
-
    __pyx_t_3 = PyObject_RichCompare(__pyx_v_n, __pyx_v_min_samples_treatment, Py_GT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 856, __pyx_L1_error)
-    __pyx_t_9 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_9 < 0)) __PYX_ERR(0, 856, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (__pyx_t_9) {
-/* … */
-      goto __pyx_L5;
-    }
-
+0857:                 p = (n_pos + parentNodeSummary[i][0] * n_reg) / (n + n_reg)
-
      __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_v_parentNodeSummary, __pyx_v_i); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_3, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_3 = PyNumber_Multiply(__pyx_t_5, __pyx_v_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      __pyx_t_5 = PyNumber_Add(__pyx_v_n_pos, __pyx_t_3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_3 = PyNumber_Add(__pyx_v_n, __pyx_v_n_reg); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 857, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_10);
-      __pyx_t_10 = 0;
-
 0858:             else:
-
+0859:                 p = parentNodeSummary[i][0]
-
    /*else*/ {
-      __pyx_t_10 = __Pyx_PyObject_GetItem(__pyx_v_parentNodeSummary, __pyx_v_i); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 859, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_10, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 859, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_3);
-      __pyx_t_3 = 0;
-    }
-    __pyx_L5:;
-
 0860:
-
+0861:             nodeSummary.append([p, n])
-
    __pyx_t_3 = PyList_New(2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 861, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_INCREF(__pyx_v_p);
-    __Pyx_GIVEREF(__pyx_v_p);
-    PyList_SET_ITEM(__pyx_t_3, 0, __pyx_v_p);
-    __Pyx_INCREF(__pyx_v_n);
-    __Pyx_GIVEREF(__pyx_v_n);
-    PyList_SET_ITEM(__pyx_t_3, 1, __pyx_v_n);
-    __pyx_t_11 = __Pyx_PyList_Append(__pyx_v_nodeSummary, __pyx_t_3); if (unlikely(__pyx_t_11 == ((int)-1))) __PYX_ERR(0, 861, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
 0862:
-
+0863:         return nodeSummary
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_nodeSummary);
-  __pyx_r = __pyx_v_nodeSummary;
-  goto __pyx_L0;
-
 0864:
-
+0865:     def uplift_classification_results(self, treatment_idx, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results[] = "\n        Classification probability for each treatment in the tree node.\n\n        Args\n        ----\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group index for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n\n        Returns\n        -------\n        res : list of list\n            The positive probabilities P(Y = 1) of each of the control and treatment groups\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results = {"uplift_classification_results", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("uplift_classification_results (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,0};
-    PyObject* values[3] = {0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, 1); __PYX_ERR(0, 865, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, 2); __PYX_ERR(0, 865, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "uplift_classification_results") < 0)) __PYX_ERR(0, 865, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 3) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_treatment_idx = values[1];
-    __pyx_v_y = values[2];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("uplift_classification_results", 1, 3, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 865, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.uplift_classification_results", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results(__pyx_self, __pyx_v_self, __pyx_v_treatment_idx, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_32uplift_classification_results(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y) {
-  PyObject *__pyx_v_counts = NULL;
-  PyObject *__pyx_v_res = NULL;
-  PyObject *__pyx_v_count = NULL;
-  PyObject *__pyx_v_n_pos = NULL;
-  PyObject *__pyx_v_n = NULL;
-  PyObject *__pyx_v_p = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("uplift_classification_results", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.uplift_classification_results", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_counts);
-  __Pyx_XDECREF(__pyx_v_res);
-  __Pyx_XDECREF(__pyx_v_count);
-  __Pyx_XDECREF(__pyx_v_n_pos);
-  __Pyx_XDECREF(__pyx_v_n);
-  __Pyx_XDECREF(__pyx_v_p);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__56 = PyTuple_Pack(9, __pyx_n_s_self, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_counts, __pyx_n_s_res, __pyx_n_s_count, __pyx_n_s_n_pos, __pyx_n_s_n, __pyx_n_s_p); if (unlikely(!__pyx_tuple__56)) __PYX_ERR(0, 865, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__56);
-  __Pyx_GIVEREF(__pyx_tuple__56);
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_33uplift_classification_results, 0, __pyx_n_s_UpliftTreeClassifier_uplift_clas, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__57)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 865, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_uplift_classification_results, __pyx_t_5) < 0) __PYX_ERR(0, 865, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_codeobj__57 = (PyObject*)__Pyx_PyCode_New(3, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__56, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_uplift_classification_results, 865, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__57)) __PYX_ERR(0, 865, __pyx_L1_error)
-
 0866:         '''
-
 0867:         Classification probability for each treatment in the tree node.
-
 0868:
-
 0869:         Args
-
 0870:         ----
-
 0871:         treatment_idx : array-like, shape = [num_samples]
-
 0872:             An array containing the treatment group index for each unit.
-
 0873:         y : array-like, shape = [num_samples]
-
 0874:             An array containing the outcome of interest for each unit.
-
 0875:
-
 0876:         Returns
-
 0877:         -------
-
 0878:         res : list of list
-
 0879:             The positive probabilities P(Y = 1) of each of the control and treatment groups
-
 0880:         '''
-
 0881:         # counts: [[N(Y=0, T=0), N(Y=1, T=0)], [N(Y=0, T=1), N(Y=1, T=1)], ...]
-
+0882:         counts = self.group_uniqueCounts(treatment_idx, y)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_group_uniqueCounts); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 882, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = NULL;
-  __pyx_t_4 = 0;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_4 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_4, 2+__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-  } else
-  #endif
-  {
-    __pyx_t_5 = PyTuple_New(2+__pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 882, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (__pyx_t_3) {
-      __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_3); __pyx_t_3 = NULL;
-    }
-    __Pyx_INCREF(__pyx_v_treatment_idx);
-    __Pyx_GIVEREF(__pyx_v_treatment_idx);
-    PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_4, __pyx_v_treatment_idx);
-    __Pyx_INCREF(__pyx_v_y);
-    __Pyx_GIVEREF(__pyx_v_y);
-    PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_4, __pyx_v_y);
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_5, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 882, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_counts = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+0883:         res = []
-
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 883, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_res = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
+0884:         for count in counts:
-
  if (likely(PyList_CheckExact(__pyx_v_counts)) || PyTuple_CheckExact(__pyx_v_counts)) {
-    __pyx_t_1 = __pyx_v_counts; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
-    __pyx_t_7 = NULL;
-  } else {
-    __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_v_counts); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 884, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 884, __pyx_L1_error)
-  }
-  for (;;) {
-    if (likely(!__pyx_t_7)) {
-      if (likely(PyList_CheckExact(__pyx_t_1))) {
-        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 884, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 884, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      } else {
-        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 884, __pyx_L1_error)
-        #else
-        __pyx_t_2 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 884, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_2);
-        #endif
-      }
-    } else {
-      __pyx_t_2 = __pyx_t_7(__pyx_t_1);
-      if (unlikely(!__pyx_t_2)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 884, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_2);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_count, __pyx_t_2);
-    __pyx_t_2 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+0885:             n_pos = count[1]
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_count, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 885, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_XDECREF_SET(__pyx_v_n_pos, __pyx_t_2);
-    __pyx_t_2 = 0;
-
+0886:             n = count[0] + n_pos
-
    __pyx_t_2 = __Pyx_GetItemInt(__pyx_v_count, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 886, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_5 = PyNumber_Add(__pyx_t_2, __pyx_v_n_pos); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 886, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_n, __pyx_t_5);
-    __pyx_t_5 = 0;
-
+0887:             p = n_pos / n if n > 0 else 0.
-
    __pyx_t_2 = PyObject_RichCompare(__pyx_v_n, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_2); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
-    __pyx_t_8 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_8 < 0)) __PYX_ERR(0, 887, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    if (__pyx_t_8) {
-      __pyx_t_2 = __Pyx_PyNumber_Divide(__pyx_v_n_pos, __pyx_v_n); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 887, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_5 = __pyx_t_2;
-      __pyx_t_2 = 0;
-    } else {
-      __Pyx_INCREF(__pyx_float_0_);
-      __pyx_t_5 = __pyx_float_0_;
-    }
-    __Pyx_XDECREF_SET(__pyx_v_p, __pyx_t_5);
-    __pyx_t_5 = 0;
-
+0888:             res.append(p)
-
    __pyx_t_9 = __Pyx_PyList_Append(__pyx_v_res, __pyx_v_p); if (unlikely(__pyx_t_9 == ((int)-1))) __PYX_ERR(0, 888, __pyx_L1_error)
-
+0889:         return res
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_res);
-  __pyx_r = __pyx_v_res;
-  goto __pyx_L0;
-
 0890:
-
+0891:     def growDecisionTreeFrom(self, X, treatment_idx, y, max_depth=10,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom[] = "\n        Train the uplift decision tree.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n        treatment_idx : array-like, shape = [num_samples]\n            An array containing the treatment group idx for each unit.\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        max_depth: int, optional (default=10)\n            The maximum depth of the tree.\n        min_samples_leaf: int, optional (default=100)\n            The minimum number of samples required to be split at a leaf node.\n        depth : int, optional (default = 1)\n            The current depth.\n        min_samples_treatment: int, optional (default=10)\n            The minimum number of samples required of the experiment group to be split at a leaf node.\n        n_reg: int, optional (default=10)\n            The regularization parameter defined in Rzepakowski et al. 2012,\n            the weight (in terms of sample size) of the parent node influence\n            on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.\n        parentNodeSummary : dictionary, optional (default = None)\n            Node summary statistics of the parent tree node.\n\n        Returns\n        -------\n        object of DecisionTree class\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom = {"growDecisionTreeFrom", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment_idx = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_max_depth = 0;
-  PyObject *__pyx_v_min_samples_leaf = 0;
-  PyObject *__pyx_v_depth = 0;
-  PyObject *__pyx_v_min_samples_treatment = 0;
-  PyObject *__pyx_v_n_reg = 0;
-  PyObject *__pyx_v_parentNodeSummary = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("growDecisionTreeFrom (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment_idx,&__pyx_n_s_y,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_depth,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_parentNodeSummary,0};
-    PyObject* values[10] = {0,0,0,0,0,0,0,0,0,0};
-    values[4] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[5] = ((PyObject *)((PyObject *)__pyx_int_100));
-    values[6] = ((PyObject *)((PyObject *)__pyx_int_1));
-    values[7] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[8] = ((PyObject *)((PyObject *)__pyx_int_100));
-/* … */
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment_idx, PyObject *__pyx_v_y, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_depth, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_parentNodeSummary) {
-  PyObject *__pyx_v_currentNodeSummary = NULL;
-  PyObject *__pyx_v_currentScore = NULL;
-  PyObject *__pyx_v_maxAbsDiff = NULL;
-  PyObject *__pyx_v_maxDiff = NULL;
-  PyObject *__pyx_v_bestTreatment = NULL;
-  PyObject *__pyx_v_suboptTreatment = NULL;
-  PyObject *__pyx_v_maxDiffTreatment = NULL;
-  PyObject *__pyx_v_maxDiffSign = NULL;
-  PyObject *__pyx_v_p_c = NULL;
-  PyObject *__pyx_v_n_c = NULL;
-  PyObject *__pyx_v_i_tr = NULL;
-  PyObject *__pyx_v_p_t = NULL;
-  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
-  PyObject *__pyx_v_diff = NULL;
-  PyObject *__pyx_v_n_t = NULL;
-  PyObject *__pyx_v_p_value = NULL;
-  PyObject *__pyx_v_upliftScore = NULL;
-  PyObject *__pyx_v_bestGain = NULL;
-  PyObject *__pyx_v_bestGainImp = NULL;
-  PyObject *__pyx_v_bestAttribute = NULL;
-  PyObject *__pyx_v_columnCount = NULL;
-  PyObject *__pyx_v_max_features = NULL;
-  PyObject *__pyx_v_col = NULL;
-  PyObject *__pyx_v_columnValues = NULL;
-  PyObject *__pyx_v_lsUnique = NULL;
-  PyObject *__pyx_v_lspercentile = NULL;
-  PyObject *__pyx_v_value = NULL;
-  PyObject *__pyx_v_X_l = NULL;
-  PyObject *__pyx_v_X_r = NULL;
-  PyObject *__pyx_v_w_l = NULL;
-  PyObject *__pyx_v_w_r = NULL;
-  PyObject *__pyx_v_y_l = NULL;
-  PyObject *__pyx_v_y_r = NULL;
-  double __pyx_v_p;
-  PyObject *__pyx_v_leftNodeSummary = NULL;
-  PyObject *__pyx_v_rightNodeSummary = NULL;
-  PyObject *__pyx_v_node_mst = NULL;
-  PyObject *__pyx_v_leftScore1 = NULL;
-  PyObject *__pyx_v_rightScore2 = NULL;
-  PyObject *__pyx_v_gain = NULL;
-  PyObject *__pyx_v_gain_for_imp = NULL;
-  PyObject *__pyx_v_n_c_left = NULL;
-  PyObject *__pyx_v_n_t_left = NULL;
-  PyObject *__pyx_v_norm_factor = NULL;
-  PyObject *__pyx_v_best_set_left = NULL;
-  PyObject *__pyx_v_best_set_right = NULL;
-  PyObject *__pyx_v_dcY = NULL;
-  PyObject *__pyx_v_i = NULL;
-  PyObject *__pyx_v_summary = NULL;
-  PyObject *__pyx_v_trueBranch = NULL;
-  PyObject *__pyx_v_falseBranch = NULL;
-  PyObject *__pyx_v_stat = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("growDecisionTreeFrom", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_XDECREF(__pyx_t_11);
-  __Pyx_XDECREF(__pyx_t_12);
-  __Pyx_XDECREF(__pyx_t_16);
-  __Pyx_XDECREF(__pyx_t_17);
-  __Pyx_XDECREF(__pyx_t_21);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.growDecisionTreeFrom", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_currentNodeSummary);
-  __Pyx_XDECREF(__pyx_v_currentScore);
-  __Pyx_XDECREF(__pyx_v_maxAbsDiff);
-  __Pyx_XDECREF(__pyx_v_maxDiff);
-  __Pyx_XDECREF(__pyx_v_bestTreatment);
-  __Pyx_XDECREF(__pyx_v_suboptTreatment);
-  __Pyx_XDECREF(__pyx_v_maxDiffTreatment);
-  __Pyx_XDECREF(__pyx_v_maxDiffSign);
-  __Pyx_XDECREF(__pyx_v_p_c);
-  __Pyx_XDECREF(__pyx_v_n_c);
-  __Pyx_XDECREF(__pyx_v_i_tr);
-  __Pyx_XDECREF(__pyx_v_p_t);
-  __Pyx_XDECREF(__pyx_v__);
-  __Pyx_XDECREF(__pyx_v_diff);
-  __Pyx_XDECREF(__pyx_v_n_t);
-  __Pyx_XDECREF(__pyx_v_p_value);
-  __Pyx_XDECREF(__pyx_v_upliftScore);
-  __Pyx_XDECREF(__pyx_v_bestGain);
-  __Pyx_XDECREF(__pyx_v_bestGainImp);
-  __Pyx_XDECREF(__pyx_v_bestAttribute);
-  __Pyx_XDECREF(__pyx_v_columnCount);
-  __Pyx_XDECREF(__pyx_v_max_features);
-  __Pyx_XDECREF(__pyx_v_col);
-  __Pyx_XDECREF(__pyx_v_columnValues);
-  __Pyx_XDECREF(__pyx_v_lsUnique);
-  __Pyx_XDECREF(__pyx_v_lspercentile);
-  __Pyx_XDECREF(__pyx_v_value);
-  __Pyx_XDECREF(__pyx_v_X_l);
-  __Pyx_XDECREF(__pyx_v_X_r);
-  __Pyx_XDECREF(__pyx_v_w_l);
-  __Pyx_XDECREF(__pyx_v_w_r);
-  __Pyx_XDECREF(__pyx_v_y_l);
-  __Pyx_XDECREF(__pyx_v_y_r);
-  __Pyx_XDECREF(__pyx_v_leftNodeSummary);
-  __Pyx_XDECREF(__pyx_v_rightNodeSummary);
-  __Pyx_XDECREF(__pyx_v_node_mst);
-  __Pyx_XDECREF(__pyx_v_leftScore1);
-  __Pyx_XDECREF(__pyx_v_rightScore2);
-  __Pyx_XDECREF(__pyx_v_gain);
-  __Pyx_XDECREF(__pyx_v_gain_for_imp);
-  __Pyx_XDECREF(__pyx_v_n_c_left);
-  __Pyx_XDECREF(__pyx_v_n_t_left);
-  __Pyx_XDECREF(__pyx_v_norm_factor);
-  __Pyx_XDECREF(__pyx_v_best_set_left);
-  __Pyx_XDECREF(__pyx_v_best_set_right);
-  __Pyx_XDECREF(__pyx_v_dcY);
-  __Pyx_XDECREF(__pyx_v_i);
-  __Pyx_XDECREF(__pyx_v_summary);
-  __Pyx_XDECREF(__pyx_v_trueBranch);
-  __Pyx_XDECREF(__pyx_v_falseBranch);
-  __Pyx_XDECREF(__pyx_v_stat);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__58 = PyTuple_Pack(63, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment_idx, __pyx_n_s_y, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_depth, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_parentNodeSummary, __pyx_n_s_currentNodeSummary, __pyx_n_s_currentScore, __pyx_n_s_maxAbsDiff, __pyx_n_s_maxDiff, __pyx_n_s_bestTreatment, __pyx_n_s_suboptTreatment, __pyx_n_s_maxDiffTreatment, __pyx_n_s_maxDiffSign, __pyx_n_s_p_c, __pyx_n_s_n_c, __pyx_n_s_i_tr, __pyx_n_s_p_t, __pyx_n_s__34, __pyx_n_s_diff, __pyx_n_s_n_t, __pyx_n_s_p_value, __pyx_n_s_upliftScore, __pyx_n_s_bestGain, __pyx_n_s_bestGainImp, __pyx_n_s_bestAttribute, __pyx_n_s_columnCount, __pyx_n_s_max_features, __pyx_n_s_col, __pyx_n_s_columnValues, __pyx_n_s_lsUnique, __pyx_n_s_lspercentile, __pyx_n_s_value, __pyx_n_s_X_l, __pyx_n_s_X_r, __pyx_n_s_w_l, __pyx_n_s_w_r, __pyx_n_s_y_l, __pyx_n_s_y_r, __pyx_n_s_p, __pyx_n_s_leftNodeSummary, __pyx_n_s_rightNodeSummary, __pyx_n_s_node_mst, __pyx_n_s_leftScore1, __pyx_n_s_rightScore2, __pyx_n_s_gain, __pyx_n_s_gain_for_imp, __pyx_n_s_n_c_left, __pyx_n_s_n_t_left, __pyx_n_s_norm_factor, __pyx_n_s_best_set_left, __pyx_n_s_best_set_right, __pyx_n_s_dcY, __pyx_n_s_i, __pyx_n_s_summary, __pyx_n_s_trueBranch, __pyx_n_s_falseBranch, __pyx_n_s_stat, __pyx_n_s_tr); if (unlikely(!__pyx_tuple__58)) __PYX_ERR(0, 891, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__58);
-  __Pyx_GIVEREF(__pyx_tuple__58);
-  __pyx_codeobj__59 = (PyObject*)__Pyx_PyCode_New(10, 0, 63, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__58, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_growDecisionTreeFrom, 891, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__59)) __PYX_ERR(0, 891, __pyx_L1_error)
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_35growDecisionTreeFrom, 0, __pyx_n_s_UpliftTreeClassifier_growDecisio, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__59)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 891, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__60);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_growDecisionTreeFrom, __pyx_t_5) < 0) __PYX_ERR(0, 891, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_tuple__60 = PyTuple_Pack(6, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_1), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_100), ((PyObject *)Py_None)); if (unlikely(!__pyx_tuple__60)) __PYX_ERR(0, 891, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__60);
-  __Pyx_GIVEREF(__pyx_tuple__60);
-
 0892:                              min_samples_leaf=100, depth=1,
-
 0893:                              min_samples_treatment=10, n_reg=100,
-
+0894:                              parentNodeSummary=None):
-
    values[9] = ((PyObject *)((PyObject *)Py_None));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 1); __PYX_ERR(0, 891, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment_idx)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 2); __PYX_ERR(0, 891, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, 3); __PYX_ERR(0, 891, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_depth);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
-          if (value) { values[7] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  8:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
-          if (value) { values[8] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  9:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_parentNodeSummary);
-          if (value) { values[9] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "growDecisionTreeFrom") < 0)) __PYX_ERR(0, 891, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment_idx = values[2];
-    __pyx_v_y = values[3];
-    __pyx_v_max_depth = values[4];
-    __pyx_v_min_samples_leaf = values[5];
-    __pyx_v_depth = values[6];
-    __pyx_v_min_samples_treatment = values[7];
-    __pyx_v_n_reg = values[8];
-    __pyx_v_parentNodeSummary = values[9];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("growDecisionTreeFrom", 0, 4, 10, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 891, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.growDecisionTreeFrom", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_34growDecisionTreeFrom(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_depth, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_parentNodeSummary);
-
 0895:         '''
-
 0896:         Train the uplift decision tree.
-
 0897:
-
 0898:         Args
-
 0899:         ----
-
 0900:         X : ndarray, shape = [num_samples, num_features]
-
 0901:             An ndarray of the covariates used to train the uplift model.
-
 0902:         treatment_idx : array-like, shape = [num_samples]
-
 0903:             An array containing the treatment group idx for each unit.
-
 0904:         y : array-like, shape = [num_samples]
-
 0905:             An array containing the outcome of interest for each unit.
-
 0906:         max_depth: int, optional (default=10)
-
 0907:             The maximum depth of the tree.
-
 0908:         min_samples_leaf: int, optional (default=100)
-
 0909:             The minimum number of samples required to be split at a leaf node.
-
 0910:         depth : int, optional (default = 1)
-
 0911:             The current depth.
-
 0912:         min_samples_treatment: int, optional (default=10)
-
 0913:             The minimum number of samples required of the experiment group to be split at a leaf node.
-
 0914:         n_reg: int, optional (default=10)
-
 0915:             The regularization parameter defined in Rzepakowski et al. 2012,
-
 0916:             the weight (in terms of sample size) of the parent node influence
-
 0917:             on the child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
-
 0918:         parentNodeSummary : dictionary, optional (default = None)
-
 0919:             Node summary statistics of the parent tree node.
-
 0920:
-
 0921:         Returns
-
 0922:         -------
-
 0923:         object of DecisionTree class
-
 0924:         '''
-
 0925:
-
+0926:         if len(X) == 0:
-
  __pyx_t_1 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_1 == ((Py_ssize_t)-1))) __PYX_ERR(0, 926, __pyx_L1_error)
-  __pyx_t_2 = ((__pyx_t_1 == 0) != 0);
-  if (__pyx_t_2) {
-/* … */
-  }
-
+0927:             return DecisionTree(classes_=self.classes_)
-
    __Pyx_XDECREF(__pyx_r);
-    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 927, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_4 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 927, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 927, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_classes, __pyx_t_5) < 0) __PYX_ERR(0, 927, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 927, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_r = __pyx_t_5;
-    __pyx_t_5 = 0;
-    goto __pyx_L0;
-
 0928:
-
 0929:         # Current node summary: [P(Y=1|T), N(T)]
-
+0930:         currentNodeSummary = self.tree_node_summary(treatment_idx, y,
-
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 930, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 930, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_INCREF(__pyx_v_treatment_idx);
-  __Pyx_GIVEREF(__pyx_v_treatment_idx);
-  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_v_treatment_idx);
-  __Pyx_INCREF(__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_v_y);
-  PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_v_y);
-/* … */
-  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_4, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 930, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_v_currentNodeSummary = __pyx_t_6;
-  __pyx_t_6 = 0;
-
+0931:                                                     min_samples_treatment=min_samples_treatment,
-
  __pyx_t_3 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 931, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
-
+0932:                                                     n_reg=n_reg,
-
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
-
+0933:                                                     parentNodeSummary=parentNodeSummary)
-
  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_parentNodeSummary, __pyx_v_parentNodeSummary) < 0) __PYX_ERR(0, 931, __pyx_L1_error)
-
+0934:         currentScore = self.evaluationFunction(currentNodeSummary)
-
  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 934, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_6 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_4, __pyx_v_currentNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_currentNodeSummary);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 934, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_v_currentScore = __pyx_t_6;
-  __pyx_t_6 = 0;
-
 0935:
-
 0936:         # Prune Stats
-
+0937:         maxAbsDiff = 0
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_maxAbsDiff = __pyx_int_0;
-
+0938:         maxDiff = -1.
-
  __Pyx_INCREF(__pyx_float_neg_1_);
-  __pyx_v_maxDiff = __pyx_float_neg_1_;
-
+0939:         bestTreatment = 0       # treatment index for the control group
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_bestTreatment = __pyx_int_0;
-
+0940:         suboptTreatment = 0     # treatment index for the control group
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_suboptTreatment = __pyx_int_0;
-
+0941:         maxDiffTreatment = 0    # treatment index for the control group
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_maxDiffTreatment = __pyx_int_0;
-
+0942:         maxDiffSign = 0
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_v_maxDiffSign = __pyx_int_0;
-
+0943:         p_c, n_c = currentNodeSummary[0]
-
  __pyx_t_6 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 943, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
-    PyObject* sequence = __pyx_t_6;
-    Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-    if (unlikely(size != 2)) {
-      if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-      else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-      __PYX_ERR(0, 943, __pyx_L1_error)
-    }
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    if (likely(PyTuple_CheckExact(sequence))) {
-      __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
-      __pyx_t_4 = PyTuple_GET_ITEM(sequence, 1); 
-    } else {
-      __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
-      __pyx_t_4 = PyList_GET_ITEM(sequence, 1); 
-    }
-    __Pyx_INCREF(__pyx_t_3);
-    __Pyx_INCREF(__pyx_t_4);
-    #else
-    __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 943, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_4 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 943, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    #endif
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  } else {
-    Py_ssize_t index = -1;
-    __pyx_t_5 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 943, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_7 = Py_TYPE(__pyx_t_5)->tp_iternext;
-    index = 0; __pyx_t_3 = __pyx_t_7(__pyx_t_5); if (unlikely(!__pyx_t_3)) goto __pyx_L4_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_3);
-    index = 1; __pyx_t_4 = __pyx_t_7(__pyx_t_5); if (unlikely(!__pyx_t_4)) goto __pyx_L4_unpacking_failed;
-    __Pyx_GOTREF(__pyx_t_4);
-    if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_5), 2) < 0) __PYX_ERR(0, 943, __pyx_L1_error)
-    __pyx_t_7 = NULL;
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    goto __pyx_L5_unpacking_done;
-    __pyx_L4_unpacking_failed:;
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_7 = NULL;
-    if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-    __PYX_ERR(0, 943, __pyx_L1_error)
-    __pyx_L5_unpacking_done:;
-  }
-  __pyx_v_p_c = __pyx_t_3;
-  __pyx_t_3 = 0;
-  __pyx_v_n_c = __pyx_t_4;
-  __pyx_t_4 = 0;
-
+0944:         for i_tr in range(1, self.n_class):
-
  __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __pyx_t_4 = PyTuple_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 944, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_INCREF(__pyx_int_1);
-  __Pyx_GIVEREF(__pyx_int_1);
-  PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_int_1);
-  __Pyx_GIVEREF(__pyx_t_6);
-  PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_t_6);
-  __pyx_t_6 = 0;
-  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_4, NULL); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_6)) || PyTuple_CheckExact(__pyx_t_6)) {
-    __pyx_t_4 = __pyx_t_6; __Pyx_INCREF(__pyx_t_4); __pyx_t_1 = 0;
-    __pyx_t_8 = NULL;
-  } else {
-    __pyx_t_1 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 944, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_8 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 944, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_8)) {
-      if (likely(PyList_CheckExact(__pyx_t_4))) {
-        if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_4)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_6 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_1); __Pyx_INCREF(__pyx_t_6); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 944, __pyx_L1_error)
-        #else
-        __pyx_t_6 = PySequence_ITEM(__pyx_t_4, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_6);
-        #endif
-      } else {
-        if (__pyx_t_1 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_6 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_1); __Pyx_INCREF(__pyx_t_6); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 944, __pyx_L1_error)
-        #else
-        __pyx_t_6 = PySequence_ITEM(__pyx_t_4, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 944, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_6);
-        #endif
-      }
-    } else {
-      __pyx_t_6 = __pyx_t_8(__pyx_t_4);
-      if (unlikely(!__pyx_t_6)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 944, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_6);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_i_tr, __pyx_t_6);
-    __pyx_t_6 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0945:             p_t, _ = currentNodeSummary[i_tr]
-
    __pyx_t_6 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_i_tr); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 945, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    if ((likely(PyTuple_CheckExact(__pyx_t_6))) || (PyList_CheckExact(__pyx_t_6))) {
-      PyObject* sequence = __pyx_t_6;
-      Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-      if (unlikely(size != 2)) {
-        if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-        else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-        __PYX_ERR(0, 945, __pyx_L1_error)
-      }
-      #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-      if (likely(PyTuple_CheckExact(sequence))) {
-        __pyx_t_3 = PyTuple_GET_ITEM(sequence, 0); 
-        __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
-      } else {
-        __pyx_t_3 = PyList_GET_ITEM(sequence, 0); 
-        __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
-      }
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_5);
-      #else
-      __pyx_t_3 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 945, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 945, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      #endif
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    } else {
-      Py_ssize_t index = -1;
-      __pyx_t_9 = PyObject_GetIter(__pyx_t_6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 945, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __pyx_t_7 = Py_TYPE(__pyx_t_9)->tp_iternext;
-      index = 0; __pyx_t_3 = __pyx_t_7(__pyx_t_9); if (unlikely(!__pyx_t_3)) goto __pyx_L8_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_3);
-      index = 1; __pyx_t_5 = __pyx_t_7(__pyx_t_9); if (unlikely(!__pyx_t_5)) goto __pyx_L8_unpacking_failed;
-      __Pyx_GOTREF(__pyx_t_5);
-      if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_9), 2) < 0) __PYX_ERR(0, 945, __pyx_L1_error)
-      __pyx_t_7 = NULL;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      goto __pyx_L9_unpacking_done;
-      __pyx_L8_unpacking_failed:;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_t_7 = NULL;
-      if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-      __PYX_ERR(0, 945, __pyx_L1_error)
-      __pyx_L9_unpacking_done:;
-    }
-    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_5);
-    __pyx_t_5 = 0;
-
 0946:             # P(Y|T=t) - P(Y|T=0)
-
+0947:             diff = p_t - p_c
-
    __pyx_t_6 = PyNumber_Subtract(__pyx_v_p_t, __pyx_v_p_c); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 947, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_XDECREF_SET(__pyx_v_diff, __pyx_t_6);
-    __pyx_t_6 = 0;
-
+0948:             if abs(diff) >= maxAbsDiff:
-
    __pyx_t_6 = __Pyx_PyNumber_Absolute(__pyx_v_diff); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 948, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_5 = PyObject_RichCompare(__pyx_t_6, __pyx_v_maxAbsDiff, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 948, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 948, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (__pyx_t_2) {
-/* … */
-    }
-
+0949:                 maxDiffTreatment = i_tr
-
      __Pyx_INCREF(__pyx_v_i_tr);
-      __Pyx_DECREF_SET(__pyx_v_maxDiffTreatment, __pyx_v_i_tr);
-
+0950:                 maxDiffSign = np.sign(diff)
-
      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 950, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_6);
-      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_sign); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 950, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __pyx_t_6 = NULL;
-      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-        __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_3);
-        if (likely(__pyx_t_6)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-          __Pyx_INCREF(__pyx_t_6);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_3, function);
-        }
-      }
-      __pyx_t_5 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_6, __pyx_v_diff) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_diff);
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 950, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_DECREF_SET(__pyx_v_maxDiffSign, __pyx_t_5);
-      __pyx_t_5 = 0;
-
+0951:                 maxAbsDiff = abs(diff)
-
      __pyx_t_5 = __Pyx_PyNumber_Absolute(__pyx_v_diff); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 951, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __Pyx_DECREF_SET(__pyx_v_maxAbsDiff, __pyx_t_5);
-      __pyx_t_5 = 0;
-
+0952:             if diff >= maxDiff:
-
    __pyx_t_5 = PyObject_RichCompare(__pyx_v_diff, __pyx_v_maxDiff, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 952, __pyx_L1_error)
-    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 952, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (__pyx_t_2) {
-/* … */
-    }
-
+0953:                 maxDiff = diff
-
      __Pyx_INCREF(__pyx_v_diff);
-      __Pyx_DECREF_SET(__pyx_v_maxDiff, __pyx_v_diff);
-
+0954:                 suboptTreatment = i_tr
-
      __Pyx_INCREF(__pyx_v_i_tr);
-      __Pyx_DECREF_SET(__pyx_v_suboptTreatment, __pyx_v_i_tr);
-
+0955:                 if diff > 0:
-
      __pyx_t_5 = PyObject_RichCompare(__pyx_v_diff, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 955, __pyx_L1_error)
-      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 955, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (__pyx_t_2) {
-/* … */
-      }
-
+0956:                     bestTreatment = i_tr
-
        __Pyx_INCREF(__pyx_v_i_tr);
-        __Pyx_DECREF_SET(__pyx_v_bestTreatment, __pyx_v_i_tr);
-
+0957:         if maxDiff > 0:
-
  __pyx_t_4 = PyObject_RichCompare(__pyx_v_maxDiff, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 957, __pyx_L1_error)
-  __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 957, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (__pyx_t_2) {
-/* … */
-    goto __pyx_L13;
-  }
-
+0958:             p_t = currentNodeSummary[bestTreatment][0]
-
    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_bestTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 958, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 958, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_5);
-    __pyx_t_5 = 0;
-
+0959:             n_t = currentNodeSummary[bestTreatment][1]
-
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_bestTreatment); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 959, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_5, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 959, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_v_n_t = __pyx_t_4;
-    __pyx_t_4 = 0;
-
 0960:         else:
-
+0961:             p_t = currentNodeSummary[suboptTreatment][0]
-
  /*else*/ {
-    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_suboptTreatment); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 961, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 961, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_p_t, __pyx_t_5);
-    __pyx_t_5 = 0;
-
+0962:             n_t = currentNodeSummary[suboptTreatment][1]
-
    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_currentNodeSummary, __pyx_v_suboptTreatment); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 962, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_4 = __Pyx_GetItemInt(__pyx_t_5, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 962, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_v_n_t = __pyx_t_4;
-    __pyx_t_4 = 0;
-  }
-  __pyx_L13:;
-
 0963:
-
+0964:         p_value = (1. - stats.norm.cdf(abs(p_c - p_t) / np.sqrt(p_t * (1 - p_t) / n_t + p_c * (1 - p_c) / n_c))) * 2
-
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_stats); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_norm); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_cdf); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = PyNumber_Subtract(__pyx_v_p_c, __pyx_v_p_t); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_6 = __Pyx_PyNumber_Absolute(__pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_sqrt); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_p_t, 1, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __pyx_t_11 = PyNumber_Multiply(__pyx_v_p_t, __pyx_t_9); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_9 = __Pyx_PyNumber_Divide(__pyx_t_11, __pyx_v_n_t); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __pyx_t_11 = __Pyx_PyInt_SubtractCObj(__pyx_int_1, __pyx_v_p_c, 1, 0, 0); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __pyx_t_12 = PyNumber_Multiply(__pyx_v_p_c, __pyx_t_11); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __pyx_t_11 = __Pyx_PyNumber_Divide(__pyx_t_12, __pyx_v_n_c); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_11);
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_t_12 = PyNumber_Add(__pyx_t_9, __pyx_t_11); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __pyx_t_11 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_10))) {
-    __pyx_t_11 = PyMethod_GET_SELF(__pyx_t_10);
-    if (likely(__pyx_t_11)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-      __Pyx_INCREF(__pyx_t_11);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_10, function);
-    }
-  }
-  __pyx_t_3 = (__pyx_t_11) ? __Pyx_PyObject_Call2Args(__pyx_t_10, __pyx_t_11, __pyx_t_12) : __Pyx_PyObject_CallOneArg(__pyx_t_10, __pyx_t_12);
-  __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = __Pyx_PyNumber_Divide(__pyx_t_6, __pyx_t_3); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_5))) {
-    __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_5);
-    if (likely(__pyx_t_3)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-      __Pyx_INCREF(__pyx_t_3);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_5, function);
-    }
-  }
-  __pyx_t_4 = (__pyx_t_3) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_3, __pyx_t_10) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_10);
-  __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyFloat_SubtractCObj(__pyx_float_1_, __pyx_t_4, 1., 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_4 = PyNumber_Multiply(__pyx_t_5, __pyx_int_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 964, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_p_value = __pyx_t_4;
-  __pyx_t_4 = 0;
-
+0965:         upliftScore = [maxDiff, p_value]
-
  __pyx_t_4 = PyList_New(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 965, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_INCREF(__pyx_v_maxDiff);
-  __Pyx_GIVEREF(__pyx_v_maxDiff);
-  PyList_SET_ITEM(__pyx_t_4, 0, __pyx_v_maxDiff);
-  __Pyx_INCREF(__pyx_v_p_value);
-  __Pyx_GIVEREF(__pyx_v_p_value);
-  PyList_SET_ITEM(__pyx_t_4, 1, __pyx_v_p_value);
-  __pyx_v_upliftScore = ((PyObject*)__pyx_t_4);
-  __pyx_t_4 = 0;
-
 0966:
-
+0967:         bestGain = 0.0
-
  __Pyx_INCREF(__pyx_float_0_0);
-  __pyx_v_bestGain = __pyx_float_0_0;
-
+0968:         bestGainImp = 0.0
-
  __Pyx_INCREF(__pyx_float_0_0);
-  __pyx_v_bestGainImp = __pyx_float_0_0;
-
+0969:         bestAttribute = None
-
  __Pyx_INCREF(Py_None);
-  __pyx_v_bestAttribute = ((PyObject*)Py_None);
-
 0970:
-
 0971:         # last column is the result/target column, 2nd to the last is the treatment group
-
+0972:         columnCount = X.shape[1]
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_X, __pyx_n_s_shape); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 972, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_5 = __Pyx_GetItemInt(__pyx_t_4, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 972, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_v_columnCount = __pyx_t_5;
-  __pyx_t_5 = 0;
-
+0973:         if (self.max_features and self.max_features > 0 and self.max_features <= columnCount):
-
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (__pyx_t_13) {
-  } else {
-    __pyx_t_2 = __pyx_t_13;
-    goto __pyx_L15_bool_binop_done;
-  }
-  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_4 = PyObject_RichCompare(__pyx_t_5, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (__pyx_t_13) {
-  } else {
-    __pyx_t_2 = __pyx_t_13;
-    goto __pyx_L15_bool_binop_done;
-  }
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_5 = PyObject_RichCompare(__pyx_t_4, __pyx_v_columnCount, Py_LE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 973, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_2 = __pyx_t_13;
-  __pyx_L15_bool_binop_done:;
-  if (__pyx_t_2) {
-/* … */
-    goto __pyx_L14;
-  }
-
+0974:             max_features = self.max_features
-
    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 974, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_v_max_features = __pyx_t_5;
-    __pyx_t_5 = 0;
-
 0975:         else:
-
+0976:             max_features = columnCount
-
  /*else*/ {
-    __Pyx_INCREF(__pyx_v_columnCount);
-    __pyx_v_max_features = __pyx_v_columnCount;
-  }
-  __pyx_L14:;
-
 0977:
-
+0978:         for col in list(self.random_state_.choice(a=range(columnCount), size=max_features, replace=False)):
-
  __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_random_state_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_choice); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_10 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_v_columnCount); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_a, __pyx_t_10) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_size, __pyx_v_max_features) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_replace, Py_False) < 0) __PYX_ERR(0, 978, __pyx_L1_error)
-  __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_5); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = PySequence_List(__pyx_t_10); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = __pyx_t_5; __Pyx_INCREF(__pyx_t_10); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  for (;;) {
-    if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_10)) break;
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_5 = PyList_GET_ITEM(__pyx_t_10, __pyx_t_1); __Pyx_INCREF(__pyx_t_5); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 978, __pyx_L1_error)
-    #else
-    __pyx_t_5 = PySequence_ITEM(__pyx_t_10, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 978, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    #endif
-    __Pyx_XDECREF_SET(__pyx_v_col, __pyx_t_5);
-    __pyx_t_5 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
+0979:             columnValues = X[:, col]
-
    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 979, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_INCREF(__pyx_slice__7);
-    __Pyx_GIVEREF(__pyx_slice__7);
-    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_slice__7);
-    __Pyx_INCREF(__pyx_v_col);
-    __Pyx_GIVEREF(__pyx_v_col);
-    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_v_col);
-    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 979, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_columnValues, __pyx_t_4);
-    __pyx_t_4 = 0;
-
 0980:             # unique values
-
+0981:             lsUnique = np.unique(columnValues)
-
    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 981, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_unique); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 981, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-      __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_3);
-      if (likely(__pyx_t_5)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-        __Pyx_INCREF(__pyx_t_5);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_3, function);
-      }
-    }
-    __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_5, __pyx_v_columnValues) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_columnValues);
-    __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 981, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_lsUnique, __pyx_t_4);
-    __pyx_t_4 = 0;
-
 0982:
-
+0983:             if np.issubdtype(lsUnique.dtype, np.number):
-
    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_np); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 983, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_t_3, __pyx_n_s_issubdtype); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 983, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_lsUnique, __pyx_n_s_dtype); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 983, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __pyx_t_6 = NULL;
-    __pyx_t_14 = 0;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_5);
-      if (likely(__pyx_t_6)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-        __Pyx_INCREF(__pyx_t_6);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_5, function);
-        __pyx_t_14 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_5)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, ((PyObject *)__pyx_ptype_5numpy_number)};
-      __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, ((PyObject *)__pyx_ptype_5numpy_number)};
-      __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    } else
-    #endif
-    {
-      __pyx_t_12 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 983, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-      if (__pyx_t_6) {
-        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_t_6); __pyx_t_6 = NULL;
-      }
-      __Pyx_GIVEREF(__pyx_t_3);
-      PyTuple_SET_ITEM(__pyx_t_12, 0+__pyx_t_14, __pyx_t_3);
-      __Pyx_INCREF(((PyObject *)__pyx_ptype_5numpy_number));
-      __Pyx_GIVEREF(((PyObject *)__pyx_ptype_5numpy_number));
-      PyTuple_SET_ITEM(__pyx_t_12, 1+__pyx_t_14, ((PyObject *)__pyx_ptype_5numpy_number));
-      __pyx_t_3 = 0;
-      __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_12, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 983, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 983, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (__pyx_t_2) {
-/* … */
-    }
-
+0984:                 if len(lsUnique) > 10:
-
      __pyx_t_15 = PyObject_Length(__pyx_v_lsUnique); if (unlikely(__pyx_t_15 == ((Py_ssize_t)-1))) __PYX_ERR(0, 984, __pyx_L1_error)
-      __pyx_t_2 = ((__pyx_t_15 > 10) != 0);
-      if (__pyx_t_2) {
-/* … */
-        goto __pyx_L21;
-      }
-
+0985:                     lspercentile = np.percentile(columnValues, [3, 5, 10, 20, 30, 50, 70, 80, 90, 95, 97])
-
        __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 985, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_percentile); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 985, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        __pyx_t_5 = PyList_New(11); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 985, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __Pyx_INCREF(__pyx_int_3);
-        __Pyx_GIVEREF(__pyx_int_3);
-        PyList_SET_ITEM(__pyx_t_5, 0, __pyx_int_3);
-        __Pyx_INCREF(__pyx_int_5);
-        __Pyx_GIVEREF(__pyx_int_5);
-        PyList_SET_ITEM(__pyx_t_5, 1, __pyx_int_5);
-        __Pyx_INCREF(__pyx_int_10);
-        __Pyx_GIVEREF(__pyx_int_10);
-        PyList_SET_ITEM(__pyx_t_5, 2, __pyx_int_10);
-        __Pyx_INCREF(__pyx_int_20);
-        __Pyx_GIVEREF(__pyx_int_20);
-        PyList_SET_ITEM(__pyx_t_5, 3, __pyx_int_20);
-        __Pyx_INCREF(__pyx_int_30);
-        __Pyx_GIVEREF(__pyx_int_30);
-        PyList_SET_ITEM(__pyx_t_5, 4, __pyx_int_30);
-        __Pyx_INCREF(__pyx_int_50);
-        __Pyx_GIVEREF(__pyx_int_50);
-        PyList_SET_ITEM(__pyx_t_5, 5, __pyx_int_50);
-        __Pyx_INCREF(__pyx_int_70);
-        __Pyx_GIVEREF(__pyx_int_70);
-        PyList_SET_ITEM(__pyx_t_5, 6, __pyx_int_70);
-        __Pyx_INCREF(__pyx_int_80);
-        __Pyx_GIVEREF(__pyx_int_80);
-        PyList_SET_ITEM(__pyx_t_5, 7, __pyx_int_80);
-        __Pyx_INCREF(__pyx_int_90);
-        __Pyx_GIVEREF(__pyx_int_90);
-        PyList_SET_ITEM(__pyx_t_5, 8, __pyx_int_90);
-        __Pyx_INCREF(__pyx_int_95);
-        __Pyx_GIVEREF(__pyx_int_95);
-        PyList_SET_ITEM(__pyx_t_5, 9, __pyx_int_95);
-        __Pyx_INCREF(__pyx_int_97);
-        __Pyx_GIVEREF(__pyx_int_97);
-        PyList_SET_ITEM(__pyx_t_5, 10, __pyx_int_97);
-        __pyx_t_3 = NULL;
-        __pyx_t_14 = 0;
-        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_12))) {
-          __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_12);
-          if (likely(__pyx_t_3)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_12);
-            __Pyx_INCREF(__pyx_t_3);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_12, function);
-            __pyx_t_14 = 1;
-          }
-        }
-        #if CYTHON_FAST_PYCALL
-        if (PyFunction_Check(__pyx_t_12)) {
-          PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_columnValues, __pyx_t_5};
-          __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_12, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        } else
-        #endif
-        #if CYTHON_FAST_PYCCALL
-        if (__Pyx_PyFastCFunction_Check(__pyx_t_12)) {
-          PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_columnValues, __pyx_t_5};
-          __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_12, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        } else
-        #endif
-        {
-          __pyx_t_6 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 985, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_6);
-          if (__pyx_t_3) {
-            __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_6, 0, __pyx_t_3); __pyx_t_3 = NULL;
-          }
-          __Pyx_INCREF(__pyx_v_columnValues);
-          __Pyx_GIVEREF(__pyx_v_columnValues);
-          PyTuple_SET_ITEM(__pyx_t_6, 0+__pyx_t_14, __pyx_v_columnValues);
-          __Pyx_GIVEREF(__pyx_t_5);
-          PyTuple_SET_ITEM(__pyx_t_6, 1+__pyx_t_14, __pyx_t_5);
-          __pyx_t_5 = 0;
-          __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_t_6, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 985, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-        }
-        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_lspercentile, __pyx_t_4);
-        __pyx_t_4 = 0;
-
 0986:                 else:
-
+0987:                     lspercentile = np.percentile(lsUnique, [10, 50, 90])
-
      /*else*/ {
-        __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_np); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 987, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_t_12, __pyx_n_s_percentile); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 987, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_6);
-        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 987, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_INCREF(__pyx_int_10);
-        __Pyx_GIVEREF(__pyx_int_10);
-        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_int_10);
-        __Pyx_INCREF(__pyx_int_50);
-        __Pyx_GIVEREF(__pyx_int_50);
-        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_int_50);
-        __Pyx_INCREF(__pyx_int_90);
-        __Pyx_GIVEREF(__pyx_int_90);
-        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_int_90);
-        __pyx_t_5 = NULL;
-        __pyx_t_14 = 0;
-        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
-          __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
-          if (likely(__pyx_t_5)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-            __Pyx_INCREF(__pyx_t_5);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_6, function);
-            __pyx_t_14 = 1;
-          }
-        }
-        #if CYTHON_FAST_PYCALL
-        if (PyFunction_Check(__pyx_t_6)) {
-          PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_lsUnique, __pyx_t_12};
-          __pyx_t_4 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-        } else
-        #endif
-        #if CYTHON_FAST_PYCCALL
-        if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
-          PyObject *__pyx_temp[3] = {__pyx_t_5, __pyx_v_lsUnique, __pyx_t_12};
-          __pyx_t_4 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
-          __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-        } else
-        #endif
-        {
-          __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 987, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          if (__pyx_t_5) {
-            __Pyx_GIVEREF(__pyx_t_5); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_5); __pyx_t_5 = NULL;
-          }
-          __Pyx_INCREF(__pyx_v_lsUnique);
-          __Pyx_GIVEREF(__pyx_v_lsUnique);
-          PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_lsUnique);
-          __Pyx_GIVEREF(__pyx_t_12);
-          PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_t_12);
-          __pyx_t_12 = 0;
-          __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_3, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 987, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        }
-        __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_lspercentile, __pyx_t_4);
-        __pyx_t_4 = 0;
-      }
-      __pyx_L21:;
-
+0988:                 lsUnique = np.unique(lspercentile)
-
      __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_np); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 988, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_6);
-      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_t_6, __pyx_n_s_unique); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 988, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __pyx_t_6 = NULL;
-      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-        __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_3);
-        if (likely(__pyx_t_6)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-          __Pyx_INCREF(__pyx_t_6);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_3, function);
-        }
-      }
-      __pyx_t_4 = (__pyx_t_6) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_6, __pyx_v_lspercentile) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_lspercentile);
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 988, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_DECREF_SET(__pyx_v_lsUnique, __pyx_t_4);
-      __pyx_t_4 = 0;
-
 0989:
-
+0990:             for value in lsUnique:
-
    if (likely(PyList_CheckExact(__pyx_v_lsUnique)) || PyTuple_CheckExact(__pyx_v_lsUnique)) {
-      __pyx_t_4 = __pyx_v_lsUnique; __Pyx_INCREF(__pyx_t_4); __pyx_t_15 = 0;
-      __pyx_t_8 = NULL;
-    } else {
-      __pyx_t_15 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_v_lsUnique); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 990, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_8 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 990, __pyx_L1_error)
-    }
-    for (;;) {
-      if (likely(!__pyx_t_8)) {
-        if (likely(PyList_CheckExact(__pyx_t_4))) {
-          if (__pyx_t_15 >= PyList_GET_SIZE(__pyx_t_4)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_3 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_15); __Pyx_INCREF(__pyx_t_3); __pyx_t_15++; if (unlikely(0 < 0)) __PYX_ERR(0, 990, __pyx_L1_error)
-          #else
-          __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_15); __pyx_t_15++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 990, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          #endif
-        } else {
-          if (__pyx_t_15 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_15); __Pyx_INCREF(__pyx_t_3); __pyx_t_15++; if (unlikely(0 < 0)) __PYX_ERR(0, 990, __pyx_L1_error)
-          #else
-          __pyx_t_3 = PySequence_ITEM(__pyx_t_4, __pyx_t_15); __pyx_t_15++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 990, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          #endif
-        }
-      } else {
-        __pyx_t_3 = __pyx_t_8(__pyx_t_4);
-        if (unlikely(!__pyx_t_3)) {
-          PyObject* exc_type = PyErr_Occurred();
-          if (exc_type) {
-            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 990, __pyx_L1_error)
-          }
-          break;
-        }
-        __Pyx_GOTREF(__pyx_t_3);
-      }
-      __Pyx_XDECREF_SET(__pyx_v_value, __pyx_t_3);
-      __pyx_t_3 = 0;
-/* … */
-      __pyx_L22_continue:;
-    }
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+0991:                 X_l, X_r, w_l, w_r, y_l, y_r = self.divideSet(X, treatment_idx, y, col, value)
-
      __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_divideSet); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 991, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_6);
-      __pyx_t_12 = NULL;
-      __pyx_t_14 = 0;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_6))) {
-        __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_6);
-        if (likely(__pyx_t_12)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-          __Pyx_INCREF(__pyx_t_12);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_6, function);
-          __pyx_t_14 = 1;
-        }
-      }
-      #if CYTHON_FAST_PYCALL
-      if (PyFunction_Check(__pyx_t_6)) {
-        PyObject *__pyx_temp[6] = {__pyx_t_12, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_col, __pyx_v_value};
-        __pyx_t_3 = __Pyx_PyFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 5+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __Pyx_GOTREF(__pyx_t_3);
-      } else
-      #endif
-      #if CYTHON_FAST_PYCCALL
-      if (__Pyx_PyFastCFunction_Check(__pyx_t_6)) {
-        PyObject *__pyx_temp[6] = {__pyx_t_12, __pyx_v_X, __pyx_v_treatment_idx, __pyx_v_y, __pyx_v_col, __pyx_v_value};
-        __pyx_t_3 = __Pyx_PyCFunction_FastCall(__pyx_t_6, __pyx_temp+1-__pyx_t_14, 5+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __Pyx_GOTREF(__pyx_t_3);
-      } else
-      #endif
-      {
-        __pyx_t_5 = PyTuple_New(5+__pyx_t_14); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 991, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        if (__pyx_t_12) {
-          __Pyx_GIVEREF(__pyx_t_12); PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_12); __pyx_t_12 = NULL;
-        }
-        __Pyx_INCREF(__pyx_v_X);
-        __Pyx_GIVEREF(__pyx_v_X);
-        PyTuple_SET_ITEM(__pyx_t_5, 0+__pyx_t_14, __pyx_v_X);
-        __Pyx_INCREF(__pyx_v_treatment_idx);
-        __Pyx_GIVEREF(__pyx_v_treatment_idx);
-        PyTuple_SET_ITEM(__pyx_t_5, 1+__pyx_t_14, __pyx_v_treatment_idx);
-        __Pyx_INCREF(__pyx_v_y);
-        __Pyx_GIVEREF(__pyx_v_y);
-        PyTuple_SET_ITEM(__pyx_t_5, 2+__pyx_t_14, __pyx_v_y);
-        __Pyx_INCREF(__pyx_v_col);
-        __Pyx_GIVEREF(__pyx_v_col);
-        PyTuple_SET_ITEM(__pyx_t_5, 3+__pyx_t_14, __pyx_v_col);
-        __Pyx_INCREF(__pyx_v_value);
-        __Pyx_GIVEREF(__pyx_v_value);
-        PyTuple_SET_ITEM(__pyx_t_5, 4+__pyx_t_14, __pyx_v_value);
-        __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_6, __pyx_t_5, NULL); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 991, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-      if ((likely(PyTuple_CheckExact(__pyx_t_3))) || (PyList_CheckExact(__pyx_t_3))) {
-        PyObject* sequence = __pyx_t_3;
-        Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-        if (unlikely(size != 6)) {
-          if (size > 6) __Pyx_RaiseTooManyValuesError(6);
-          else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-          __PYX_ERR(0, 991, __pyx_L1_error)
-        }
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        if (likely(PyTuple_CheckExact(sequence))) {
-          __pyx_t_6 = PyTuple_GET_ITEM(sequence, 0); 
-          __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
-          __pyx_t_12 = PyTuple_GET_ITEM(sequence, 2); 
-          __pyx_t_11 = PyTuple_GET_ITEM(sequence, 3); 
-          __pyx_t_9 = PyTuple_GET_ITEM(sequence, 4); 
-          __pyx_t_16 = PyTuple_GET_ITEM(sequence, 5); 
-        } else {
-          __pyx_t_6 = PyList_GET_ITEM(sequence, 0); 
-          __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
-          __pyx_t_12 = PyList_GET_ITEM(sequence, 2); 
-          __pyx_t_11 = PyList_GET_ITEM(sequence, 3); 
-          __pyx_t_9 = PyList_GET_ITEM(sequence, 4); 
-          __pyx_t_16 = PyList_GET_ITEM(sequence, 5); 
-        }
-        __Pyx_INCREF(__pyx_t_6);
-        __Pyx_INCREF(__pyx_t_5);
-        __Pyx_INCREF(__pyx_t_12);
-        __Pyx_INCREF(__pyx_t_11);
-        __Pyx_INCREF(__pyx_t_9);
-        __Pyx_INCREF(__pyx_t_16);
-        #else
-        {
-          Py_ssize_t i;
-          PyObject** temps[6] = {&__pyx_t_6,&__pyx_t_5,&__pyx_t_12,&__pyx_t_11,&__pyx_t_9,&__pyx_t_16};
-          for (i=0; i < 6; i++) {
-            PyObject* item = PySequence_ITEM(sequence, i); if (unlikely(!item)) __PYX_ERR(0, 991, __pyx_L1_error)
-            __Pyx_GOTREF(item);
-            *(temps[i]) = item;
-          }
-        }
-        #endif
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      } else {
-        Py_ssize_t index = -1;
-        PyObject** temps[6] = {&__pyx_t_6,&__pyx_t_5,&__pyx_t_12,&__pyx_t_11,&__pyx_t_9,&__pyx_t_16};
-        __pyx_t_17 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_17)) __PYX_ERR(0, 991, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_17);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_7 = Py_TYPE(__pyx_t_17)->tp_iternext;
-        for (index=0; index < 6; index++) {
-          PyObject* item = __pyx_t_7(__pyx_t_17); if (unlikely(!item)) goto __pyx_L24_unpacking_failed;
-          __Pyx_GOTREF(item);
-          *(temps[index]) = item;
-        }
-        if (__Pyx_IternextUnpackEndCheck(__pyx_t_7(__pyx_t_17), 6) < 0) __PYX_ERR(0, 991, __pyx_L1_error)
-        __pyx_t_7 = NULL;
-        __Pyx_DECREF(__pyx_t_17); __pyx_t_17 = 0;
-        goto __pyx_L25_unpacking_done;
-        __pyx_L24_unpacking_failed:;
-        __Pyx_DECREF(__pyx_t_17); __pyx_t_17 = 0;
-        __pyx_t_7 = NULL;
-        if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-        __PYX_ERR(0, 991, __pyx_L1_error)
-        __pyx_L25_unpacking_done:;
-      }
-      __Pyx_XDECREF_SET(__pyx_v_X_l, __pyx_t_6);
-      __pyx_t_6 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_X_r, __pyx_t_5);
-      __pyx_t_5 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_w_l, __pyx_t_12);
-      __pyx_t_12 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_w_r, __pyx_t_11);
-      __pyx_t_11 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_y_l, __pyx_t_9);
-      __pyx_t_9 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_y_r, __pyx_t_16);
-      __pyx_t_16 = 0;
-
 0992:                 # check the split validity on min_samples_leaf  372
-
+0993:                 if (len(X_l) < min_samples_leaf or len(X_r) < min_samples_leaf):
-
      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 993, __pyx_L1_error)
-      __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_16 = PyObject_RichCompare(__pyx_t_3, __pyx_v_min_samples_leaf, Py_LT); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      if (!__pyx_t_13) {
-      } else {
-        __pyx_t_2 = __pyx_t_13;
-        goto __pyx_L27_bool_binop_done;
-      }
-      __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 993, __pyx_L1_error)
-      __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __pyx_t_3 = PyObject_RichCompare(__pyx_t_16, __pyx_v_min_samples_leaf, Py_LT); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 993, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_2 = __pyx_t_13;
-      __pyx_L27_bool_binop_done:;
-      if (__pyx_t_2) {
-/* … */
-      }
-
+0994:                     continue
-
        goto __pyx_L22_continue;
-
 0995:                 # summarize notes
-
 0996:                 # Gain -- Entropy or Gini
-
+0997:                 p = float(len(X_l)) / len(X)
-
      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 997, __pyx_L1_error)
-      __pyx_t_19 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_19 == ((Py_ssize_t)-1))) __PYX_ERR(0, 997, __pyx_L1_error)
-      __pyx_v_p = (((double)__pyx_t_18) / __pyx_t_19);
-
+0998:                 leftNodeSummary = self.tree_node_summary(w_l, y_l,
-
      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 998, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_16 = PyTuple_New(2); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 998, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __Pyx_INCREF(__pyx_v_w_l);
-      __Pyx_GIVEREF(__pyx_v_w_l);
-      PyTuple_SET_ITEM(__pyx_t_16, 0, __pyx_v_w_l);
-      __Pyx_INCREF(__pyx_v_y_l);
-      __Pyx_GIVEREF(__pyx_v_y_l);
-      PyTuple_SET_ITEM(__pyx_t_16, 1, __pyx_v_y_l);
-/* … */
-      __pyx_t_11 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 998, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_leftNodeSummary, __pyx_t_11);
-      __pyx_t_11 = 0;
-
+0999:                                                          min_samples_treatment=min_samples_treatment,
-
      __pyx_t_9 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 999, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
-
+1000:                                                          n_reg=n_reg,
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
-
+1001:                                                          parentNodeSummary=currentNodeSummary)
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 999, __pyx_L1_error)
-
 1002:
-
+1003:                 rightNodeSummary = self.tree_node_summary(w_r, y_r,
-
      __pyx_t_11 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_tree_node_summary); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1003, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_11);
-      __pyx_t_9 = PyTuple_New(2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1003, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __Pyx_INCREF(__pyx_v_w_r);
-      __Pyx_GIVEREF(__pyx_v_w_r);
-      PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_v_w_r);
-      __Pyx_INCREF(__pyx_v_y_r);
-      __Pyx_GIVEREF(__pyx_v_y_r);
-      PyTuple_SET_ITEM(__pyx_t_9, 1, __pyx_v_y_r);
-/* … */
-      __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_11, __pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1003, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_rightNodeSummary, __pyx_t_3);
-      __pyx_t_3 = 0;
-
+1004:                                                           min_samples_treatment=min_samples_treatment,
-
      __pyx_t_16 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1004, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
-
+1005:                                                           n_reg=n_reg,
-
      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
-
+1006:                                                           parentNodeSummary=currentNodeSummary)
-
      if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1004, __pyx_L1_error)
-
 1007:
-
 1008:                 # check the split validity on min_samples_treatment
-
+1009:                 assert len(leftNodeSummary) == len(rightNodeSummary)
-
      #ifndef CYTHON_WITHOUT_ASSERTIONS
-      if (unlikely(!Py_OptimizeFlag)) {
-        __pyx_t_19 = PyObject_Length(__pyx_v_leftNodeSummary); if (unlikely(__pyx_t_19 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1009, __pyx_L1_error)
-        __pyx_t_18 = PyObject_Length(__pyx_v_rightNodeSummary); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1009, __pyx_L1_error)
-        if (unlikely(!((__pyx_t_19 == __pyx_t_18) != 0))) {
-          PyErr_SetNone(PyExc_AssertionError);
-          __PYX_ERR(0, 1009, __pyx_L1_error)
-        }
-      }
-      #endif
-
 1010:
-
+1011:                 node_mst = min([stat[1] for stat in leftNodeSummary + rightNodeSummary])
-
      __pyx_t_3 = PyList_New(0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1011, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_16 = PyNumber_Add(__pyx_v_leftNodeSummary, __pyx_v_rightNodeSummary); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      if (likely(PyList_CheckExact(__pyx_t_16)) || PyTuple_CheckExact(__pyx_t_16)) {
-        __pyx_t_9 = __pyx_t_16; __Pyx_INCREF(__pyx_t_9); __pyx_t_18 = 0;
-        __pyx_t_20 = NULL;
-      } else {
-        __pyx_t_18 = -1; __pyx_t_9 = PyObject_GetIter(__pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1011, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_20 = Py_TYPE(__pyx_t_9)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1011, __pyx_L1_error)
-      }
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      for (;;) {
-        if (likely(!__pyx_t_20)) {
-          if (likely(PyList_CheckExact(__pyx_t_9))) {
-            if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_9)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_16 = PyList_GET_ITEM(__pyx_t_9, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1011, __pyx_L1_error)
-            #else
-            __pyx_t_16 = PySequence_ITEM(__pyx_t_9, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_16);
-            #endif
-          } else {
-            if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_9)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_16 = PyTuple_GET_ITEM(__pyx_t_9, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1011, __pyx_L1_error)
-            #else
-            __pyx_t_16 = PySequence_ITEM(__pyx_t_9, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_16);
-            #endif
-          }
-        } else {
-          __pyx_t_16 = __pyx_t_20(__pyx_t_9);
-          if (unlikely(!__pyx_t_16)) {
-            PyObject* exc_type = PyErr_Occurred();
-            if (exc_type) {
-              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-              else __PYX_ERR(0, 1011, __pyx_L1_error)
-            }
-            break;
-          }
-          __Pyx_GOTREF(__pyx_t_16);
-        }
-        __Pyx_XDECREF_SET(__pyx_v_stat, __pyx_t_16);
-        __pyx_t_16 = 0;
-        __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_stat, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1011, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        if (unlikely(__Pyx_ListComp_Append(__pyx_t_3, (PyObject*)__pyx_t_16))) __PYX_ERR(0, 1011, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_t_9 = __Pyx_PyObject_CallOneArg(__pyx_builtin_min, __pyx_t_3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1011, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __Pyx_XDECREF_SET(__pyx_v_node_mst, __pyx_t_9);
-      __pyx_t_9 = 0;
-
+1012:                 if node_mst < min_samples_treatment:
-
      __pyx_t_9 = PyObject_RichCompare(__pyx_v_node_mst, __pyx_v_min_samples_treatment, Py_LT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1012, __pyx_L1_error)
-      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1012, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      if (__pyx_t_2) {
-/* … */
-      }
-
+1013:                     continue
-
        goto __pyx_L22_continue;
-
 1014:
-
 1015:                 # evaluate the split
-
+1016:                 if self.evaluationFunction == self.evaluate_CTS:
-
      __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1016, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1016, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      __pyx_t_16 = PyObject_RichCompare(__pyx_t_9, __pyx_t_3, Py_EQ); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1016, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1016, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      if (__pyx_t_2) {
-/* … */
-        goto __pyx_L32;
-      }
-
+1017:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
-
        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1017, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_9 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
-          __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_3);
-          if (likely(__pyx_t_9)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-            __Pyx_INCREF(__pyx_t_9);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_3, function);
-          }
-        }
-        __pyx_t_16 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_9, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_leftNodeSummary);
-        __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-        if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1017, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_16);
-        __pyx_t_16 = 0;
-
+1018:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
-
        __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1018, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_9 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_3))) {
-          __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_3);
-          if (likely(__pyx_t_9)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-            __Pyx_INCREF(__pyx_t_9);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_3, function);
-          }
-        }
-        __pyx_t_16 = (__pyx_t_9) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_9, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_v_rightNodeSummary);
-        __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-        if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1018, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_16);
-        __pyx_t_16 = 0;
-
+1019:                     gain = (currentScore - p * leftScore1 - (1 - p) * rightScore2)
-
        __pyx_t_16 = PyFloat_FromDouble(__pyx_v_p); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_3 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_16 = PyNumber_Subtract(__pyx_v_currentScore, __pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = PyFloat_FromDouble((1.0 - __pyx_v_p)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = PyNumber_Subtract(__pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1019, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1020:                     gain_for_imp = (len(X) * currentScore - len(X_l) * leftScore1 - len(X_r) * rightScore2)
-
        __pyx_t_18 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_currentScore); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_16 = PyNumber_Multiply(__pyx_t_3, __pyx_v_leftScore1); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = PyNumber_Subtract(__pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_16 = PyNumber_Subtract(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1020, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_16);
-        __pyx_t_16 = 0;
-
+1021:                 elif self.evaluationFunction == self.evaluate_DDP:
-
      __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1021, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_DDP); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1021, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_3 = PyObject_RichCompare(__pyx_t_16, __pyx_t_9, Py_EQ); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1021, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1021, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      if (__pyx_t_2) {
-/* … */
-        goto __pyx_L32;
-      }
-
+1022:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
-
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1022, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_16 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
-          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
-          if (likely(__pyx_t_16)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
-            __Pyx_INCREF(__pyx_t_16);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_9, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_leftNodeSummary);
-        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1022, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1023:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
-
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1023, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_16 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
-          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
-          if (likely(__pyx_t_16)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
-            __Pyx_INCREF(__pyx_t_16);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_9, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_rightNodeSummary);
-        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1023, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1024:                     gain = np.abs(leftScore1 - rightScore2)
-
        __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_np); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1024, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_t_9, __pyx_n_s_abs); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1024, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __pyx_t_9 = PyNumber_Subtract(__pyx_v_leftScore1, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1024, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_11 = NULL;
-        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_16))) {
-          __pyx_t_11 = PyMethod_GET_SELF(__pyx_t_16);
-          if (likely(__pyx_t_11)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_16);
-            __Pyx_INCREF(__pyx_t_11);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_16, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_11) ? __Pyx_PyObject_Call2Args(__pyx_t_16, __pyx_t_11, __pyx_t_9) : __Pyx_PyObject_CallOneArg(__pyx_t_16, __pyx_t_9);
-        __Pyx_XDECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1024, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1025:                     gain_for_imp = np.abs(len(X_l) * leftScore1 - len(X_r) * rightScore2)
-
        __Pyx_GetModuleGlobalName(__pyx_t_16, __pyx_n_s_np); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_t_16, __pyx_n_s_abs); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_11 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_11)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_11);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_12 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_16 = PyNumber_Subtract(__pyx_t_11, __pyx_t_12); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_11); __pyx_t_11 = 0;
-        __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __pyx_t_12 = NULL;
-        if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_9))) {
-          __pyx_t_12 = PyMethod_GET_SELF(__pyx_t_9);
-          if (likely(__pyx_t_12)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
-            __Pyx_INCREF(__pyx_t_12);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_9, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_12) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_12, __pyx_t_16) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_t_16);
-        __Pyx_XDECREF(__pyx_t_12); __pyx_t_12 = 0;
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1025, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_3);
-        __pyx_t_3 = 0;
-
 1026:                 else:
-
+1027:                     leftScore1 = self.evaluationFunction(leftNodeSummary)
-
      /*else*/ {
-        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1027, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_16 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
-          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
-          if (likely(__pyx_t_16)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
-            __Pyx_INCREF(__pyx_t_16);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_9, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_leftNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_leftNodeSummary);
-        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1027, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_leftScore1, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1028:                     rightScore2 = self.evaluationFunction(rightNodeSummary)
-
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1028, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_16 = NULL;
-        if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_9))) {
-          __pyx_t_16 = PyMethod_GET_SELF(__pyx_t_9);
-          if (likely(__pyx_t_16)) {
-            PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_9);
-            __Pyx_INCREF(__pyx_t_16);
-            __Pyx_INCREF(function);
-            __Pyx_DECREF_SET(__pyx_t_9, function);
-          }
-        }
-        __pyx_t_3 = (__pyx_t_16) ? __Pyx_PyObject_Call2Args(__pyx_t_9, __pyx_t_16, __pyx_v_rightNodeSummary) : __Pyx_PyObject_CallOneArg(__pyx_t_9, __pyx_v_rightNodeSummary);
-        __Pyx_XDECREF(__pyx_t_16); __pyx_t_16 = 0;
-        if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1028, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_rightScore2, __pyx_t_3);
-        __pyx_t_3 = 0;
-
+1029:                     gain = (p * leftScore1 + (1 - p) * rightScore2 - currentScore)
-
        __pyx_t_3 = PyFloat_FromDouble(__pyx_v_p); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_t_3, __pyx_v_leftScore1); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = PyFloat_FromDouble((1.0 - __pyx_v_p)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __pyx_t_16 = PyNumber_Multiply(__pyx_t_3, __pyx_v_rightScore2); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __pyx_t_3 = PyNumber_Add(__pyx_t_9, __pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_16 = PyNumber_Subtract(__pyx_t_3, __pyx_v_currentScore); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1029, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain, __pyx_t_16);
-        __pyx_t_16 = 0;
-
+1030:                     gain_for_imp = (len(X_l) * leftScore1 + len(X_r) * rightScore2 - len(X) * currentScore)
-
        __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_3 = PyNumber_Multiply(__pyx_t_16, __pyx_v_leftScore1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_t_16, __pyx_v_rightScore2); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __pyx_t_16 = PyNumber_Add(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __pyx_t_18 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __pyx_t_9 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_3 = PyNumber_Multiply(__pyx_t_9, __pyx_v_currentScore); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __pyx_t_9 = PyNumber_Subtract(__pyx_t_16, __pyx_t_3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1030, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_XDECREF_SET(__pyx_v_gain_for_imp, __pyx_t_9);
-        __pyx_t_9 = 0;
-
+1031:                     if self.normalization:
-
        __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_normalization); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1031, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1031, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        if (__pyx_t_2) {
-/* … */
-          goto __pyx_L33;
-        }
-
+1032:                         n_c = currentNodeSummary[0][1]
-
          __pyx_t_9 = __Pyx_GetItemInt(__pyx_v_currentNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1032, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __pyx_t_3 = __Pyx_GetItemInt(__pyx_t_9, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1032, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-          __Pyx_DECREF_SET(__pyx_v_n_c, __pyx_t_3);
-          __pyx_t_3 = 0;
-
+1033:                         n_c_left = leftNodeSummary[0][1]
-
          __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_leftNodeSummary, 0, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1033, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          __pyx_t_9 = __Pyx_GetItemInt(__pyx_t_3, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1033, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-          __Pyx_XDECREF_SET(__pyx_v_n_c_left, __pyx_t_9);
-          __pyx_t_9 = 0;
-
+1034:                         n_t = [tr[1] for tr in currentNodeSummary[1:]]
-
          __pyx_t_9 = PyList_New(0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1034, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __pyx_t_3 = __Pyx_PyObject_GetSlice(__pyx_v_currentNodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
-            __pyx_t_16 = __pyx_t_3; __Pyx_INCREF(__pyx_t_16); __pyx_t_18 = 0;
-            __pyx_t_20 = NULL;
-          } else {
-            __pyx_t_18 = -1; __pyx_t_16 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1034, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_16);
-            __pyx_t_20 = Py_TYPE(__pyx_t_16)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1034, __pyx_L1_error)
-          }
-          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-          for (;;) {
-            if (likely(!__pyx_t_20)) {
-              if (likely(PyList_CheckExact(__pyx_t_16))) {
-                if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_16)) break;
-                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-                __pyx_t_3 = PyList_GET_ITEM(__pyx_t_16, __pyx_t_18); __Pyx_INCREF(__pyx_t_3); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1034, __pyx_L1_error)
-                #else
-                __pyx_t_3 = PySequence_ITEM(__pyx_t_16, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
-                __Pyx_GOTREF(__pyx_t_3);
-                #endif
-              } else {
-                if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_16)) break;
-                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-                __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_16, __pyx_t_18); __Pyx_INCREF(__pyx_t_3); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1034, __pyx_L1_error)
-                #else
-                __pyx_t_3 = PySequence_ITEM(__pyx_t_16, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
-                __Pyx_GOTREF(__pyx_t_3);
-                #endif
-              }
-            } else {
-              __pyx_t_3 = __pyx_t_20(__pyx_t_16);
-              if (unlikely(!__pyx_t_3)) {
-                PyObject* exc_type = PyErr_Occurred();
-                if (exc_type) {
-                  if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-                  else __PYX_ERR(0, 1034, __pyx_L1_error)
-                }
-                break;
-              }
-              __Pyx_GOTREF(__pyx_t_3);
-            }
-            __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_3);
-            __pyx_t_3 = 0;
-            __pyx_t_3 = __Pyx_GetItemInt(__pyx_v_tr, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1034, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_3);
-            if (unlikely(__Pyx_ListComp_Append(__pyx_t_9, (PyObject*)__pyx_t_3))) __PYX_ERR(0, 1034, __pyx_L1_error)
-            __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-          }
-          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-          __Pyx_DECREF_SET(__pyx_v_n_t, __pyx_t_9);
-          __pyx_t_9 = 0;
-
+1035:                         n_t_left = [tr[1] for tr in leftNodeSummary[1:]]
-
          __pyx_t_9 = PyList_New(0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1035, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __pyx_t_16 = __Pyx_PyObject_GetSlice(__pyx_v_leftNodeSummary, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_16);
-          if (likely(PyList_CheckExact(__pyx_t_16)) || PyTuple_CheckExact(__pyx_t_16)) {
-            __pyx_t_3 = __pyx_t_16; __Pyx_INCREF(__pyx_t_3); __pyx_t_18 = 0;
-            __pyx_t_20 = NULL;
-          } else {
-            __pyx_t_18 = -1; __pyx_t_3 = PyObject_GetIter(__pyx_t_16); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1035, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_3);
-            __pyx_t_20 = Py_TYPE(__pyx_t_3)->tp_iternext; if (unlikely(!__pyx_t_20)) __PYX_ERR(0, 1035, __pyx_L1_error)
-          }
-          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-          for (;;) {
-            if (likely(!__pyx_t_20)) {
-              if (likely(PyList_CheckExact(__pyx_t_3))) {
-                if (__pyx_t_18 >= PyList_GET_SIZE(__pyx_t_3)) break;
-                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-                __pyx_t_16 = PyList_GET_ITEM(__pyx_t_3, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1035, __pyx_L1_error)
-                #else
-                __pyx_t_16 = PySequence_ITEM(__pyx_t_3, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
-                __Pyx_GOTREF(__pyx_t_16);
-                #endif
-              } else {
-                if (__pyx_t_18 >= PyTuple_GET_SIZE(__pyx_t_3)) break;
-                #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-                __pyx_t_16 = PyTuple_GET_ITEM(__pyx_t_3, __pyx_t_18); __Pyx_INCREF(__pyx_t_16); __pyx_t_18++; if (unlikely(0 < 0)) __PYX_ERR(0, 1035, __pyx_L1_error)
-                #else
-                __pyx_t_16 = PySequence_ITEM(__pyx_t_3, __pyx_t_18); __pyx_t_18++; if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
-                __Pyx_GOTREF(__pyx_t_16);
-                #endif
-              }
-            } else {
-              __pyx_t_16 = __pyx_t_20(__pyx_t_3);
-              if (unlikely(!__pyx_t_16)) {
-                PyObject* exc_type = PyErr_Occurred();
-                if (exc_type) {
-                  if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-                  else __PYX_ERR(0, 1035, __pyx_L1_error)
-                }
-                break;
-              }
-              __Pyx_GOTREF(__pyx_t_16);
-            }
-            __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_16);
-            __pyx_t_16 = 0;
-            __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_tr, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1035, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_16);
-            if (unlikely(__Pyx_ListComp_Append(__pyx_t_9, (PyObject*)__pyx_t_16))) __PYX_ERR(0, 1035, __pyx_L1_error)
-            __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-          }
-          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-          __Pyx_XDECREF_SET(__pyx_v_n_t_left, ((PyObject*)__pyx_t_9));
-          __pyx_t_9 = 0;
-
 1036:
-
+1037:                         norm_factor = self.normI(n_c, n_c_left, n_t, n_t_left, alpha=0.9)
-
          __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_normI); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1037, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __pyx_t_3 = PyTuple_New(4); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1037, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_3);
-          __Pyx_INCREF(__pyx_v_n_c);
-          __Pyx_GIVEREF(__pyx_v_n_c);
-          PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_v_n_c);
-          __Pyx_INCREF(__pyx_v_n_c_left);
-          __Pyx_GIVEREF(__pyx_v_n_c_left);
-          PyTuple_SET_ITEM(__pyx_t_3, 1, __pyx_v_n_c_left);
-          __Pyx_INCREF(__pyx_v_n_t);
-          __Pyx_GIVEREF(__pyx_v_n_t);
-          PyTuple_SET_ITEM(__pyx_t_3, 2, __pyx_v_n_t);
-          __Pyx_INCREF(__pyx_v_n_t_left);
-          __Pyx_GIVEREF(__pyx_v_n_t_left);
-          PyTuple_SET_ITEM(__pyx_t_3, 3, __pyx_v_n_t_left);
-          __pyx_t_16 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1037, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_16);
-          if (PyDict_SetItem(__pyx_t_16, __pyx_n_s_alpha, __pyx_float_0_9) < 0) __PYX_ERR(0, 1037, __pyx_L1_error)
-          __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_9, __pyx_t_3, __pyx_t_16); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1037, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_12);
-          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-          __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-          __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-          __Pyx_XDECREF_SET(__pyx_v_norm_factor, __pyx_t_12);
-          __pyx_t_12 = 0;
-
 1038:                     else:
-
+1039:                         norm_factor = 1
-
        /*else*/ {
-          __Pyx_INCREF(__pyx_int_1);
-          __Pyx_XDECREF_SET(__pyx_v_norm_factor, __pyx_int_1);
-        }
-        __pyx_L33:;
-
+1040:                     gain = gain / norm_factor
-
        __pyx_t_12 = __Pyx_PyNumber_Divide(__pyx_v_gain, __pyx_v_norm_factor); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1040, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_DECREF_SET(__pyx_v_gain, __pyx_t_12);
-        __pyx_t_12 = 0;
-      }
-      __pyx_L32:;
-
+1041:                 if (gain > bestGain and len(X_l) > min_samples_leaf and len(X_r) > min_samples_leaf):
-
      __pyx_t_12 = PyObject_RichCompare(__pyx_v_gain, __pyx_v_bestGain, Py_GT); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      if (__pyx_t_13) {
-      } else {
-        __pyx_t_2 = __pyx_t_13;
-        goto __pyx_L39_bool_binop_done;
-      }
-      __pyx_t_18 = PyObject_Length(__pyx_v_X_l); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __pyx_t_12 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-      __pyx_t_16 = PyObject_RichCompare(__pyx_t_12, __pyx_v_min_samples_leaf, Py_GT); __Pyx_XGOTREF(__pyx_t_16); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_16); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      if (__pyx_t_13) {
-      } else {
-        __pyx_t_2 = __pyx_t_13;
-        goto __pyx_L39_bool_binop_done;
-      }
-      __pyx_t_18 = PyObject_Length(__pyx_v_X_r); if (unlikely(__pyx_t_18 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __pyx_t_16 = PyInt_FromSsize_t(__pyx_t_18); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __pyx_t_12 = PyObject_RichCompare(__pyx_t_16, __pyx_v_min_samples_leaf, Py_GT); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1041, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      __pyx_t_2 = __pyx_t_13;
-      __pyx_L39_bool_binop_done:;
-      if (__pyx_t_2) {
-/* … */
-      }
-
+1042:                     bestGain = gain
-
        __Pyx_INCREF(__pyx_v_gain);
-        __Pyx_DECREF_SET(__pyx_v_bestGain, __pyx_v_gain);
-
+1043:                     bestGainImp = gain_for_imp
-
        __Pyx_INCREF(__pyx_v_gain_for_imp);
-        __Pyx_DECREF_SET(__pyx_v_bestGainImp, __pyx_v_gain_for_imp);
-
+1044:                     bestAttribute = (col, value)
-
        __pyx_t_12 = PyTuple_New(2); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1044, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_INCREF(__pyx_v_col);
-        __Pyx_GIVEREF(__pyx_v_col);
-        PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_v_col);
-        __Pyx_INCREF(__pyx_v_value);
-        __Pyx_GIVEREF(__pyx_v_value);
-        PyTuple_SET_ITEM(__pyx_t_12, 1, __pyx_v_value);
-        __Pyx_DECREF_SET(__pyx_v_bestAttribute, ((PyObject*)__pyx_t_12));
-        __pyx_t_12 = 0;
-
+1045:                     best_set_left = [X_l, w_l, y_l]
-
        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1045, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_INCREF(__pyx_v_X_l);
-        __Pyx_GIVEREF(__pyx_v_X_l);
-        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_v_X_l);
-        __Pyx_INCREF(__pyx_v_w_l);
-        __Pyx_GIVEREF(__pyx_v_w_l);
-        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_v_w_l);
-        __Pyx_INCREF(__pyx_v_y_l);
-        __Pyx_GIVEREF(__pyx_v_y_l);
-        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_v_y_l);
-        __Pyx_XDECREF_SET(__pyx_v_best_set_left, ((PyObject*)__pyx_t_12));
-        __pyx_t_12 = 0;
-
+1046:                     best_set_right = [X_r, w_r, y_r]
-
        __pyx_t_12 = PyList_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1046, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_INCREF(__pyx_v_X_r);
-        __Pyx_GIVEREF(__pyx_v_X_r);
-        PyList_SET_ITEM(__pyx_t_12, 0, __pyx_v_X_r);
-        __Pyx_INCREF(__pyx_v_w_r);
-        __Pyx_GIVEREF(__pyx_v_w_r);
-        PyList_SET_ITEM(__pyx_t_12, 1, __pyx_v_w_r);
-        __Pyx_INCREF(__pyx_v_y_r);
-        __Pyx_GIVEREF(__pyx_v_y_r);
-        PyList_SET_ITEM(__pyx_t_12, 2, __pyx_v_y_r);
-        __Pyx_XDECREF_SET(__pyx_v_best_set_right, ((PyObject*)__pyx_t_12));
-        __pyx_t_12 = 0;
-
 1047:
-
+1048:         dcY = {'impurity': '%.3f' % currentScore, 'samples': '%d' % len(X)}
-
  __pyx_t_10 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __pyx_t_4 = __Pyx_PyString_FormatSafe(__pyx_kp_s_3f, __pyx_v_currentScore); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  if (PyDict_SetItem(__pyx_t_10, __pyx_n_s_impurity, __pyx_t_4) < 0) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __pyx_t_1 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_1 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_12 = __Pyx_PyString_Format(__pyx_kp_s_d, __pyx_t_4); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (PyDict_SetItem(__pyx_t_10, __pyx_n_s_samples, __pyx_t_12) < 0) __PYX_ERR(0, 1048, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __pyx_v_dcY = ((PyObject*)__pyx_t_10);
-  __pyx_t_10 = 0;
-
 1049:         # Add treatment size
-
+1050:         dcY['group_size'] = ''
-
  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_group_size, __pyx_kp_s__4) < 0)) __PYX_ERR(0, 1050, __pyx_L1_error)
-
+1051:         for i, summary in enumerate(currentNodeSummary):
-
  __Pyx_INCREF(__pyx_int_0);
-  __pyx_t_10 = __pyx_int_0;
-  if (likely(PyList_CheckExact(__pyx_v_currentNodeSummary)) || PyTuple_CheckExact(__pyx_v_currentNodeSummary)) {
-    __pyx_t_12 = __pyx_v_currentNodeSummary; __Pyx_INCREF(__pyx_t_12); __pyx_t_1 = 0;
-    __pyx_t_8 = NULL;
-  } else {
-    __pyx_t_1 = -1; __pyx_t_12 = PyObject_GetIter(__pyx_v_currentNodeSummary); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1051, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __pyx_t_8 = Py_TYPE(__pyx_t_12)->tp_iternext; if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1051, __pyx_L1_error)
-  }
-  for (;;) {
-    if (likely(!__pyx_t_8)) {
-      if (likely(PyList_CheckExact(__pyx_t_12))) {
-        if (__pyx_t_1 >= PyList_GET_SIZE(__pyx_t_12)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_12, __pyx_t_1); __Pyx_INCREF(__pyx_t_4); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 1051, __pyx_L1_error)
-        #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_12, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        #endif
-      } else {
-        if (__pyx_t_1 >= PyTuple_GET_SIZE(__pyx_t_12)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_12, __pyx_t_1); __Pyx_INCREF(__pyx_t_4); __pyx_t_1++; if (unlikely(0 < 0)) __PYX_ERR(0, 1051, __pyx_L1_error)
-        #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_12, __pyx_t_1); __pyx_t_1++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        #endif
-      }
-    } else {
-      __pyx_t_4 = __pyx_t_8(__pyx_t_12);
-      if (unlikely(!__pyx_t_4)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1051, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_4);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_summary, __pyx_t_4);
-    __pyx_t_4 = 0;
-    __Pyx_INCREF(__pyx_t_10);
-    __Pyx_XDECREF_SET(__pyx_v_i, __pyx_t_10);
-    __pyx_t_4 = __Pyx_PyInt_AddObjC(__pyx_t_10, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1051, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_10);
-    __pyx_t_10 = __pyx_t_4;
-    __pyx_t_4 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
+1052:             dcY['group_size'] += ' ' + self.classes_[i] + ': ' + str(summary[1])
-
    __Pyx_INCREF(__pyx_n_s_group_size);
-    __pyx_t_21 = __pyx_n_s_group_size;
-    __pyx_t_4 = __Pyx_PyDict_GetItem(__pyx_v_dcY, __pyx_t_21); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __pyx_t_3 = __Pyx_PyObject_GetItem(__pyx_t_16, __pyx_v_i); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __pyx_t_16 = PyNumber_Add(__pyx_kp_s__5, __pyx_t_3); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __pyx_t_3 = PyNumber_Add(__pyx_t_16, __pyx_kp_s__6); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __pyx_t_16 = __Pyx_GetItemInt(__pyx_v_summary, 1, long, 1, __Pyx_PyInt_From_long, 0, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __pyx_t_9 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyString_Type)), __pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __pyx_t_16 = PyNumber_Add(__pyx_t_3, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_4, __pyx_t_16); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_t_21, __pyx_t_9) < 0)) __PYX_ERR(0, 1052, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __Pyx_DECREF(__pyx_t_21); __pyx_t_21 = 0;
-
+1053:         dcY['upliftScore'] = [round(upliftScore[0], 4), round(upliftScore[1], 4)]
-
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  __Pyx_INCREF(__pyx_int_4);
-  __Pyx_GIVEREF(__pyx_int_4);
-  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
-  __pyx_t_12 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_12);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 1));
-  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 1));
-  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 1));
-  __Pyx_INCREF(__pyx_int_4);
-  __Pyx_GIVEREF(__pyx_int_4);
-  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
-  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  __pyx_t_10 = PyList_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_GIVEREF(__pyx_t_12);
-  PyList_SET_ITEM(__pyx_t_10, 0, __pyx_t_12);
-  __Pyx_GIVEREF(__pyx_t_9);
-  PyList_SET_ITEM(__pyx_t_10, 1, __pyx_t_9);
-  __pyx_t_12 = 0;
-  __pyx_t_9 = 0;
-  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_upliftScore, __pyx_t_10) < 0)) __PYX_ERR(0, 1053, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-
+1054:         dcY['matchScore'] = round(upliftScore[0], 4)
-
  __pyx_t_10 = PyTuple_New(2); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1054, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_10);
-  __Pyx_INCREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  __Pyx_GIVEREF(PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  PyTuple_SET_ITEM(__pyx_t_10, 0, PyList_GET_ITEM(__pyx_v_upliftScore, 0));
-  __Pyx_INCREF(__pyx_int_4);
-  __Pyx_GIVEREF(__pyx_int_4);
-  PyTuple_SET_ITEM(__pyx_t_10, 1, __pyx_int_4);
-  __pyx_t_9 = __Pyx_PyObject_Call(__pyx_builtin_round, __pyx_t_10, NULL); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1054, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_9);
-  __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-  if (unlikely(PyDict_SetItem(__pyx_v_dcY, __pyx_n_s_matchScore, __pyx_t_9) < 0)) __PYX_ERR(0, 1054, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-
 1055:
-
+1056:         if bestGain > 0 and depth < max_depth:
-
  __pyx_t_9 = PyObject_RichCompare(__pyx_v_bestGain, __pyx_int_0, Py_GT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1056, __pyx_L1_error)
-  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  if (__pyx_t_13) {
-  } else {
-    __pyx_t_2 = __pyx_t_13;
-    goto __pyx_L45_bool_binop_done;
-  }
-  __pyx_t_9 = PyObject_RichCompare(__pyx_v_depth, __pyx_v_max_depth, Py_LT); __Pyx_XGOTREF(__pyx_t_9); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1056, __pyx_L1_error)
-  __pyx_t_13 = __Pyx_PyObject_IsTrue(__pyx_t_9); if (unlikely(__pyx_t_13 < 0)) __PYX_ERR(0, 1056, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-  __pyx_t_2 = __pyx_t_13;
-  __pyx_L45_bool_binop_done:;
-  if (__pyx_t_2) {
-/* … */
-  }
-
+1057:             self.feature_imp_dict[bestAttribute[0]] += bestGainImp
-
    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_feature_imp_dict); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1057, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 1057, __pyx_L1_error)
-    }
-    __Pyx_INCREF(PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0));
-    __pyx_t_10 = PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0);
-    __pyx_t_12 = __Pyx_PyObject_GetItem(__pyx_t_9, __pyx_t_10); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1057, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __pyx_t_16 = PyNumber_InPlaceAdd(__pyx_t_12, __pyx_v_bestGainImp); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1057, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    if (unlikely(PyObject_SetItem(__pyx_t_9, __pyx_t_10, __pyx_t_16) < 0)) __PYX_ERR(0, 1057, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-
+1058:             trueBranch = self.growDecisionTreeFrom(
-
    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1058, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-/* … */
-    __pyx_t_10 = PySequence_Tuple(__pyx_v_best_set_left); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1058, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-/* … */
-    __pyx_t_12 = PyTuple_New(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1058, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __Pyx_INCREF(__pyx_v_max_depth);
-    __Pyx_GIVEREF(__pyx_v_max_depth);
-    PyTuple_SET_ITEM(__pyx_t_12, 0, __pyx_v_max_depth);
-    __Pyx_INCREF(__pyx_v_min_samples_leaf);
-    __Pyx_GIVEREF(__pyx_v_min_samples_leaf);
-    PyTuple_SET_ITEM(__pyx_t_12, 1, __pyx_v_min_samples_leaf);
-    __Pyx_GIVEREF(__pyx_t_16);
-    PyTuple_SET_ITEM(__pyx_t_12, 2, __pyx_t_16);
-    __pyx_t_16 = 0;
-    __pyx_t_16 = PyNumber_Add(__pyx_t_10, __pyx_t_12); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1058, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-/* … */
-    __pyx_t_10 = __Pyx_PyObject_Call(__pyx_t_9, __pyx_t_16, __pyx_t_12); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1058, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __pyx_v_trueBranch = __pyx_t_10;
-    __pyx_t_10 = 0;
-
+1059:                 *best_set_left, max_depth, min_samples_leaf,
-
    if (unlikely(!__pyx_v_best_set_left)) { __Pyx_RaiseUnboundLocalError("best_set_left"); __PYX_ERR(0, 1059, __pyx_L1_error) }
-
+1060:                 depth + 1, min_samples_treatment=min_samples_treatment,
-
    __pyx_t_16 = __Pyx_PyInt_AddObjC(__pyx_v_depth, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1060, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-/* … */
-    __pyx_t_12 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1060, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
-
+1061:                 n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_12, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1060, __pyx_L1_error)
-
 1062:             )
-
+1063:             falseBranch = self.growDecisionTreeFrom(
-
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_growDecisionTreeFrom); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1063, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-/* … */
-    __pyx_t_12 = PySequence_Tuple(__pyx_v_best_set_right); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1063, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-/* … */
-    __pyx_t_9 = PyTuple_New(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1063, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __Pyx_INCREF(__pyx_v_max_depth);
-    __Pyx_GIVEREF(__pyx_v_max_depth);
-    PyTuple_SET_ITEM(__pyx_t_9, 0, __pyx_v_max_depth);
-    __Pyx_INCREF(__pyx_v_min_samples_leaf);
-    __Pyx_GIVEREF(__pyx_v_min_samples_leaf);
-    PyTuple_SET_ITEM(__pyx_t_9, 1, __pyx_v_min_samples_leaf);
-    __Pyx_GIVEREF(__pyx_t_16);
-    PyTuple_SET_ITEM(__pyx_t_9, 2, __pyx_t_16);
-    __pyx_t_16 = 0;
-    __pyx_t_16 = PyNumber_Add(__pyx_t_12, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1063, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-/* … */
-    __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_16, __pyx_t_9); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1063, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_v_falseBranch = __pyx_t_12;
-    __pyx_t_12 = 0;
-
+1064:                 *best_set_right, max_depth, min_samples_leaf,
-
    if (unlikely(!__pyx_v_best_set_right)) { __Pyx_RaiseUnboundLocalError("best_set_right"); __PYX_ERR(0, 1064, __pyx_L1_error) }
-
+1065:                 depth + 1, min_samples_treatment=min_samples_treatment,
-
    __pyx_t_16 = __Pyx_PyInt_AddObjC(__pyx_v_depth, __pyx_int_1, 1, 0, 0); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1065, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-/* … */
-    __pyx_t_9 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1065, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
-
+1066:                 n_reg=n_reg, parentNodeSummary=currentNodeSummary
-
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_parentNodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1065, __pyx_L1_error)
-
 1067:             )
-
 1068:
-
+1069:             return DecisionTree(
-
    __Pyx_XDECREF(__pyx_r);
-    __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1069, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_12);
-/* … */
-    __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1069, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_r = __pyx_t_16;
-    __pyx_t_16 = 0;
-    goto __pyx_L0;
-
+1070:                 classes_=self.classes_,
-
    __pyx_t_9 = __Pyx_PyDict_NewPresized(12); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1070, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1070, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_16) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-
+1071:                 col=bestAttribute[0], value=bestAttribute[1],
-
    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 1071, __pyx_L1_error)
-    }
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_col, PyTuple_GET_ITEM(__pyx_v_bestAttribute, 0)) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    if (unlikely(__pyx_v_bestAttribute == Py_None)) {
-      PyErr_SetString(PyExc_TypeError, "'NoneType' object is not subscriptable");
-      __PYX_ERR(0, 1071, __pyx_L1_error)
-    }
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_value, PyTuple_GET_ITEM(__pyx_v_bestAttribute, 1)) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-
+1072:                 trueBranch=trueBranch, falseBranch=falseBranch, summary=dcY,
-
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_trueBranch, __pyx_v_trueBranch) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_falseBranch, __pyx_v_falseBranch) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-
+1073:                 maxDiffTreatment=maxDiffTreatment, maxDiffSign=maxDiffSign,
-
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-
+1074:                 nodeSummary=currentNodeSummary,
-
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-
+1075:                 backupResults=self.uplift_classification_results(treatment_idx, y),
-
    __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1075, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_10);
-    __pyx_t_4 = NULL;
-    __pyx_t_14 = 0;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-      __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_10);
-      if (likely(__pyx_t_4)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-        __Pyx_INCREF(__pyx_t_4);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_10, function);
-        __pyx_t_14 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_10)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
-      __pyx_t_16 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_GOTREF(__pyx_t_16);
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-      PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
-      __pyx_t_16 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_GOTREF(__pyx_t_16);
-    } else
-    #endif
-    {
-      __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1075, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_3);
-      if (__pyx_t_4) {
-        __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_4); __pyx_t_4 = NULL;
-      }
-      __Pyx_INCREF(__pyx_v_treatment_idx);
-      __Pyx_GIVEREF(__pyx_v_treatment_idx);
-      PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_treatment_idx);
-      __Pyx_INCREF(__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_v_y);
-      __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, NULL); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1075, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_backupResults, __pyx_t_16) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-
+1076:                 bestTreatment=bestTreatment, upliftScore=upliftScore
-
    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-    if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1070, __pyx_L1_error)
-
 1077:             )
-
 1078:         else:
-
+1079:             if self.evaluationFunction == self.evaluate_CTS:
-
  /*else*/ {
-    __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1079, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_16);
-    __pyx_t_9 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_evaluate_CTS); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1079, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __pyx_t_12 = PyObject_RichCompare(__pyx_t_16, __pyx_t_9, Py_EQ); __Pyx_XGOTREF(__pyx_t_12); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1079, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_12); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1079, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-    if (__pyx_t_2) {
-/* … */
-    }
-
+1080:                 return DecisionTree(
-
      __Pyx_XDECREF(__pyx_r);
-      __Pyx_GetModuleGlobalName(__pyx_t_12, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1080, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-/* … */
-      __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_12, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1080, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_r = __pyx_t_16;
-      __pyx_t_16 = 0;
-      goto __pyx_L0;
-
+1081:                     classes_=self.classes_,
-
      __pyx_t_9 = __Pyx_PyDict_NewPresized(6); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1081, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_16 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1081, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_16) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-
+1082:                     results=self.uplift_classification_results(treatment_idx, y),
-
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1082, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __pyx_t_3 = NULL;
-      __pyx_t_14 = 0;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-        __pyx_t_3 = PyMethod_GET_SELF(__pyx_t_10);
-        if (likely(__pyx_t_3)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-          __Pyx_INCREF(__pyx_t_3);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_10, function);
-          __pyx_t_14 = 1;
-        }
-      }
-      #if CYTHON_FAST_PYCALL
-      if (PyFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-        __pyx_t_16 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_GOTREF(__pyx_t_16);
-      } else
-      #endif
-      #if CYTHON_FAST_PYCCALL
-      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[3] = {__pyx_t_3, __pyx_v_treatment_idx, __pyx_v_y};
-        __pyx_t_16 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_3); __pyx_t_3 = 0;
-        __Pyx_GOTREF(__pyx_t_16);
-      } else
-      #endif
-      {
-        __pyx_t_4 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1082, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        if (__pyx_t_3) {
-          __Pyx_GIVEREF(__pyx_t_3); PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_t_3); __pyx_t_3 = NULL;
-        }
-        __Pyx_INCREF(__pyx_v_treatment_idx);
-        __Pyx_GIVEREF(__pyx_v_treatment_idx);
-        PyTuple_SET_ITEM(__pyx_t_4, 0+__pyx_t_14, __pyx_v_treatment_idx);
-        __Pyx_INCREF(__pyx_v_y);
-        __Pyx_GIVEREF(__pyx_v_y);
-        PyTuple_SET_ITEM(__pyx_t_4, 1+__pyx_t_14, __pyx_v_y);
-        __pyx_t_16 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_4, NULL); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1082, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_16);
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_results, __pyx_t_16) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-
+1083:                     summary=dcY, nodeSummary=currentNodeSummary,
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-
+1084:                     bestTreatment=bestTreatment, upliftScore=upliftScore
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1081, __pyx_L1_error)
-
 1085:                 )
-
 1086:             else:
-
+1087:                 return DecisionTree(
-
    /*else*/ {
-      __Pyx_XDECREF(__pyx_r);
-      __Pyx_GetModuleGlobalName(__pyx_t_16, __pyx_n_s_DecisionTree); if (unlikely(!__pyx_t_16)) __PYX_ERR(0, 1087, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_16);
-/* … */
-      __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_16, __pyx_empty_tuple, __pyx_t_9); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1087, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-      __Pyx_DECREF(__pyx_t_16); __pyx_t_16 = 0;
-      __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-      __pyx_r = __pyx_t_12;
-      __pyx_t_12 = 0;
-      goto __pyx_L0;
-    }
-  }
-
+1088:                     classes_=self.classes_,
-
      __pyx_t_9 = __Pyx_PyDict_NewPresized(8); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1088, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_9);
-      __pyx_t_12 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1088, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_12);
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_classes, __pyx_t_12) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-
+1089:                     results=self.uplift_classification_results(treatment_idx, y),
-
      __pyx_t_10 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_uplift_classification_results); if (unlikely(!__pyx_t_10)) __PYX_ERR(0, 1089, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_10);
-      __pyx_t_4 = NULL;
-      __pyx_t_14 = 0;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_10))) {
-        __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_10);
-        if (likely(__pyx_t_4)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_10);
-          __Pyx_INCREF(__pyx_t_4);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_10, function);
-          __pyx_t_14 = 1;
-        }
-      }
-      #if CYTHON_FAST_PYCALL
-      if (PyFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
-        __pyx_t_12 = __Pyx_PyFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_GOTREF(__pyx_t_12);
-      } else
-      #endif
-      #if CYTHON_FAST_PYCCALL
-      if (__Pyx_PyFastCFunction_Check(__pyx_t_10)) {
-        PyObject *__pyx_temp[3] = {__pyx_t_4, __pyx_v_treatment_idx, __pyx_v_y};
-        __pyx_t_12 = __Pyx_PyCFunction_FastCall(__pyx_t_10, __pyx_temp+1-__pyx_t_14, 2+__pyx_t_14); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
-        __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __Pyx_GOTREF(__pyx_t_12);
-      } else
-      #endif
-      {
-        __pyx_t_3 = PyTuple_New(2+__pyx_t_14); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1089, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        if (__pyx_t_4) {
-          __Pyx_GIVEREF(__pyx_t_4); PyTuple_SET_ITEM(__pyx_t_3, 0, __pyx_t_4); __pyx_t_4 = NULL;
-        }
-        __Pyx_INCREF(__pyx_v_treatment_idx);
-        __Pyx_GIVEREF(__pyx_v_treatment_idx);
-        PyTuple_SET_ITEM(__pyx_t_3, 0+__pyx_t_14, __pyx_v_treatment_idx);
-        __Pyx_INCREF(__pyx_v_y);
-        __Pyx_GIVEREF(__pyx_v_y);
-        PyTuple_SET_ITEM(__pyx_t_3, 1+__pyx_t_14, __pyx_v_y);
-        __pyx_t_12 = __Pyx_PyObject_Call(__pyx_t_10, __pyx_t_3, NULL); if (unlikely(!__pyx_t_12)) __PYX_ERR(0, 1089, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_12);
-        __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-      }
-      __Pyx_DECREF(__pyx_t_10); __pyx_t_10 = 0;
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_results, __pyx_t_12) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_12); __pyx_t_12 = 0;
-
+1090:                     summary=dcY, maxDiffTreatment=maxDiffTreatment,
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_summary, __pyx_v_dcY) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffTreatment, __pyx_v_maxDiffTreatment) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-
+1091:                     maxDiffSign=maxDiffSign, nodeSummary=currentNodeSummary,
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_maxDiffSign, __pyx_v_maxDiffSign) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_nodeSummary, __pyx_v_currentNodeSummary) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-
+1092:                     bestTreatment=bestTreatment, upliftScore=upliftScore
-
      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_bestTreatment, __pyx_v_bestTreatment) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-      if (PyDict_SetItem(__pyx_t_9, __pyx_n_s_upliftScore, __pyx_v_upliftScore) < 0) __PYX_ERR(0, 1088, __pyx_L1_error)
-
 1093:                 )
-
 1094:
-
+1095:     @staticmethod
-
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1095, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_classify, __pyx_t_2) < 0) __PYX_ERR(0, 1096, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1096:     def classify(observations, tree, dataMissing=False):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify[] = "\n        Classifies (prediction) the observations according to the tree.\n\n        Args\n        ----\n        observations : list of list\n            The internal data format for the training data (combining X, Y, treatment).\n\n        dataMissing: boolean, optional (default = False)\n            An indicator for if data are missing or not.\n\n        Returns\n        -------\n        tree.results, tree.upliftScore :\n            The results in the leaf node.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify = {"classify", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_observations = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_v_dataMissing = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classify (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,&__pyx_n_s_dataMissing,0};
-    PyObject* values[3] = {0,0,0};
-    values[2] = ((PyObject *)((PyObject *)Py_False));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("classify", 0, 2, 3, 1); __PYX_ERR(0, 1096, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_dataMissing);
-          if (value) { values[2] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classify") < 0)) __PYX_ERR(0, 1096, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_observations = values[0];
-    __pyx_v_tree = values[1];
-    __pyx_v_dataMissing = values[2];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("classify", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1096, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify(__pyx_self, __pyx_v_observations, __pyx_v_tree, __pyx_v_dataMissing);
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_36classify(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree, PyObject *__pyx_v_dataMissing) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classify", 0);
-  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify, __pyx_empty_tuple, NULL);
-  if (unlikely(!__pyx_cur_scope)) {
-    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *)Py_None);
-    __Pyx_INCREF(Py_None);
-    __PYX_ERR(0, 1096, __pyx_L1_error)
-  } else {
-    __Pyx_GOTREF(__pyx_cur_scope);
-  }
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__61 = PyTuple_Pack(7, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_dataMissing, __pyx_n_s_classifyWithoutMissingData, __pyx_n_s_classifyWithoutMissingData, __pyx_n_s_classifyWithMissingData, __pyx_n_s_classifyWithMissingData); if (unlikely(!__pyx_tuple__61)) __PYX_ERR(0, 1096, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__61);
-  __Pyx_GIVEREF(__pyx_tuple__61);
-  __pyx_codeobj__62 = (PyObject*)__Pyx_PyCode_New(3, 0, 7, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__61, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classify, 1096, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__62)) __PYX_ERR(0, 1096, __pyx_L1_error)
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_37classify, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftTreeClassifier_classify, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__62)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1096, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__63);
-  __pyx_tuple__63 = PyTuple_Pack(1, ((PyObject *)Py_False)); if (unlikely(!__pyx_tuple__63)) __PYX_ERR(0, 1096, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__63);
-  __Pyx_GIVEREF(__pyx_tuple__63);
-/* … */
-struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify {
-  PyObject_HEAD
-  PyObject *__pyx_v_classifyWithMissingData;
-  PyObject *__pyx_v_classifyWithoutMissingData;
-};
-
-
 1097:         '''
-
 1098:         Classifies (prediction) the observations according to the tree.
-
 1099:
-
 1100:         Args
-
 1101:         ----
-
 1102:         observations : list of list
-
 1103:             The internal data format for the training data (combining X, Y, treatment).
-
 1104:
-
 1105:         dataMissing: boolean, optional (default = False)
-
 1106:             An indicator for if data are missing or not.
-
 1107:
-
 1108:         Returns
-
 1109:         -------
-
 1110:         tree.results, tree.upliftScore :
-
 1111:             The results in the leaf node.
-
 1112:         '''
-
 1113:
-
+1114:         def classifyWithoutMissingData(observations, tree):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData[] = "\n            Classifies (prediction) the observations according to the tree, assuming without missing data.\n\n            Args\n            ----\n            observations : list of list\n                The internal data format for the training data (combining X, Y, treatment).\n\n            Returns\n            -------\n            tree.results, tree.upliftScore :\n                The results in the leaf node.\n            ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData = {"classifyWithoutMissingData", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_observations = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classifyWithoutMissingData (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,0};
-    PyObject* values[2] = {0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("classifyWithoutMissingData", 1, 2, 2, 1); __PYX_ERR(0, 1114, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classifyWithoutMissingData") < 0)) __PYX_ERR(0, 1114, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-    }
-    __pyx_v_observations = values[0];
-    __pyx_v_tree = values[1];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("classifyWithoutMissingData", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1114, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithoutMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_self, __pyx_v_observations, __pyx_v_tree);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_outer_scope;
-  PyObject *__pyx_v_v = NULL;
-  PyObject *__pyx_v_branch = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classifyWithoutMissingData", 0);
-  __pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *) __Pyx_CyFunction_GetClosure(__pyx_self);
-  __pyx_cur_scope = __pyx_outer_scope;
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithoutMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_v);
-  __Pyx_XDECREF(__pyx_v_branch);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__8 = PyTuple_Pack(4, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_v, __pyx_n_s_branch); if (unlikely(!__pyx_tuple__8)) __PYX_ERR(0, 1114, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__8);
-  __Pyx_GIVEREF(__pyx_tuple__8);
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_1classifyWithoutMissingData, 0, __pyx_n_s_UpliftTreeClassifier_classify_lo, ((PyObject*)__pyx_cur_scope), __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__9)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1114, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GIVEREF(__pyx_t_1);
-  __pyx_cur_scope->__pyx_v_classifyWithoutMissingData = __pyx_t_1;
-  __pyx_t_1 = 0;
-  __pyx_codeobj__9 = (PyObject*)__Pyx_PyCode_New(2, 0, 4, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__8, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classifyWithoutMissingData, 1114, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__9)) __PYX_ERR(0, 1114, __pyx_L1_error)
-
 1115:             '''
-
 1116:             Classifies (prediction) the observations according to the tree, assuming without missing data.
-
 1117:
-
 1118:             Args
-
 1119:             ----
-
 1120:             observations : list of list
-
 1121:                 The internal data format for the training data (combining X, Y, treatment).
-
 1122:
-
 1123:             Returns
-
 1124:             -------
-
 1125:             tree.results, tree.upliftScore :
-
 1126:                 The results in the leaf node.
-
 1127:             '''
-
+1128:             if tree.results is not None:  # leaf
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1128, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = (__pyx_t_1 != Py_None);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_3 = (__pyx_t_2 != 0);
-  if (__pyx_t_3) {
-/* … */
-  }
-
+1129:                 return tree.results, tree.upliftScore
-
    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1129, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_upliftScore); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1129, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = PyTuple_New(2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1129, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_GIVEREF(__pyx_t_1);
-    PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_1);
-    __Pyx_GIVEREF(__pyx_t_4);
-    PyTuple_SET_ITEM(__pyx_t_5, 1, __pyx_t_4);
-    __pyx_t_1 = 0;
-    __pyx_t_4 = 0;
-    __pyx_r = __pyx_t_5;
-    __pyx_t_5 = 0;
-    goto __pyx_L0;
-
 1130:             else:
-
+1131:                 v = observations[tree.col]
-
  /*else*/ {
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1131, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_observations, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1131, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_v_v = __pyx_t_4;
-    __pyx_t_4 = 0;
-
+1132:                 branch = None
-
    __Pyx_INCREF(Py_None);
-    __pyx_v_branch = Py_None;
-
+1133:                 if isinstance(v, int) or isinstance(v, float):
-
    __pyx_t_2 = PyInt_Check(__pyx_v_v); 
-    __pyx_t_6 = (__pyx_t_2 != 0);
-    if (!__pyx_t_6) {
-    } else {
-      __pyx_t_3 = __pyx_t_6;
-      goto __pyx_L5_bool_binop_done;
-    }
-    __pyx_t_6 = PyFloat_Check(__pyx_v_v); 
-    __pyx_t_2 = (__pyx_t_6 != 0);
-    __pyx_t_3 = __pyx_t_2;
-    __pyx_L5_bool_binop_done:;
-    if (__pyx_t_3) {
-/* … */
-      goto __pyx_L4;
-    }
-
+1134:                     if v >= tree.value:
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1134, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_5 = PyObject_RichCompare(__pyx_v_v, __pyx_t_4, Py_GE); __Pyx_XGOTREF(__pyx_t_5); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1134, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_5); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1134, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (__pyx_t_3) {
-/* … */
-        goto __pyx_L7;
-      }
-
+1135:                         branch = tree.trueBranch
-
        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1135, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_5);
-        __pyx_t_5 = 0;
-
 1136:                     else:
-
+1137:                         branch = tree.falseBranch
-
      /*else*/ {
-        __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1137, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_5);
-        __pyx_t_5 = 0;
-      }
-      __pyx_L7:;
-
 1138:                 else:
-
+1139:                     if v == tree.value:
-
    /*else*/ {
-      __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1139, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __pyx_t_4 = PyObject_RichCompare(__pyx_v_v, __pyx_t_5, Py_EQ); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1139, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1139, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      if (__pyx_t_3) {
-/* … */
-        goto __pyx_L8;
-      }
-
+1140:                         branch = tree.trueBranch
-
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1140, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
-        __pyx_t_4 = 0;
-
 1141:                     else:
-
+1142:                         branch = tree.falseBranch
-
      /*else*/ {
-        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1142, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
-        __pyx_t_4 = 0;
-      }
-      __pyx_L8:;
-    }
-    __pyx_L4:;
-  }
-
+1143:             return classifyWithoutMissingData(observations, branch)
-
  __Pyx_XDECREF(__pyx_r);
-  if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithoutMissingData)) { __Pyx_RaiseClosureNameError("classifyWithoutMissingData"); __PYX_ERR(0, 1143, __pyx_L1_error) }
-  __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_cur_scope->__pyx_v_classifyWithoutMissingData, __pyx_v_observations, __pyx_v_branch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1143, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_r = __pyx_t_4;
-  __pyx_t_4 = 0;
-  goto __pyx_L0;
-
 1144:
-
+1145:         def classifyWithMissingData(observations, tree):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData[] = "\n            Classifies (prediction) the observations according to the tree, assuming with missing data.\n\n            Args\n            ----\n            observations : list of list\n                The internal data format for the training data (combining X, Y, treatment).\n\n            Returns\n            -------\n            tree.results, tree.upliftScore :\n                The results in the leaf node.\n            ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData = {"classifyWithMissingData", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_observations = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classifyWithMissingData (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_observations,&__pyx_n_s_tree,0};
-    PyObject* values[2] = {0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_observations)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("classifyWithMissingData", 1, 2, 2, 1); __PYX_ERR(0, 1145, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "classifyWithMissingData") < 0)) __PYX_ERR(0, 1145, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 2) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-    }
-    __pyx_v_observations = values[0];
-    __pyx_v_tree = values[1];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("classifyWithMissingData", 1, 2, 2, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1145, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_self, __pyx_v_observations, __pyx_v_tree);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(PyObject *__pyx_self, PyObject *__pyx_v_observations, PyObject *__pyx_v_tree) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_cur_scope;
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *__pyx_outer_scope;
-  PyObject *__pyx_v_v = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_v_fr = NULL;
-  PyObject *__pyx_v_tcount = NULL;
-  PyObject *__pyx_v_fcount = NULL;
-  PyObject *__pyx_v_tw = NULL;
-  PyObject *__pyx_v_fw = NULL;
-  PyObject *__pyx_v_result = NULL;
-  PyObject *__pyx_v_k = NULL;
-  PyObject *__pyx_v_branch = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("classifyWithMissingData", 0);
-  __pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct__classify *) __Pyx_CyFunction_GetClosure(__pyx_self);
-  __pyx_cur_scope = __pyx_outer_scope;
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftTreeClassifier.classify.classifyWithMissingData", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_v);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XDECREF(__pyx_v_fr);
-  __Pyx_XDECREF(__pyx_v_tcount);
-  __Pyx_XDECREF(__pyx_v_fcount);
-  __Pyx_XDECREF(__pyx_v_tw);
-  __Pyx_XDECREF(__pyx_v_fw);
-  __Pyx_XDECREF(__pyx_v_result);
-  __Pyx_XDECREF(__pyx_v_k);
-  __Pyx_XDECREF(__pyx_v_branch);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__10 = PyTuple_Pack(12, __pyx_n_s_observations, __pyx_n_s_tree, __pyx_n_s_v, __pyx_n_s_tr, __pyx_n_s_fr, __pyx_n_s_tcount, __pyx_n_s_fcount, __pyx_n_s_tw, __pyx_n_s_fw, __pyx_n_s_result, __pyx_n_s_k, __pyx_n_s_branch); if (unlikely(!__pyx_tuple__10)) __PYX_ERR(0, 1145, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__10);
-  __Pyx_GIVEREF(__pyx_tuple__10);
-/* … */
-  __pyx_t_1 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_3classifyWithMissingData, 0, __pyx_n_s_UpliftTreeClassifier_classify_lo_2, ((PyObject*)__pyx_cur_scope), __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__11)); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1145, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GIVEREF(__pyx_t_1);
-  __pyx_cur_scope->__pyx_v_classifyWithMissingData = __pyx_t_1;
-  __pyx_t_1 = 0;
-  __pyx_codeobj__11 = (PyObject*)__Pyx_PyCode_New(2, 0, 12, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__10, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_classifyWithMissingData, 1145, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__11)) __PYX_ERR(0, 1145, __pyx_L1_error)
-
 1146:             '''
-
 1147:             Classifies (prediction) the observations according to the tree, assuming with missing data.
-
 1148:
-
 1149:             Args
-
 1150:             ----
-
 1151:             observations : list of list
-
 1152:                 The internal data format for the training data (combining X, Y, treatment).
-
 1153:
-
 1154:             Returns
-
 1155:             -------
-
 1156:             tree.results, tree.upliftScore :
-
 1157:                 The results in the leaf node.
-
 1158:             '''
-
+1159:             if tree.results is not None:  # leaf
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1159, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = (__pyx_t_1 != Py_None);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_3 = (__pyx_t_2 != 0);
-  if (__pyx_t_3) {
-/* … */
-  }
-
+1160:                 return tree.results
-
    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_results); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1160, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_r = __pyx_t_1;
-    __pyx_t_1 = 0;
-    goto __pyx_L0;
-
 1161:             else:
-
+1162:                 v = observations[tree.col]
-
  /*else*/ {
-    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_col); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1162, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = __Pyx_PyObject_GetItem(__pyx_v_observations, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1162, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_v_v = __pyx_t_4;
-    __pyx_t_4 = 0;
-
+1163:                 if v is None:
-
    __pyx_t_3 = (__pyx_v_v == Py_None);
-    __pyx_t_2 = (__pyx_t_3 != 0);
-    if (__pyx_t_2) {
-/* … */
-    }
-
+1164:                     tr = classifyWithMissingData(observations, tree.trueBranch)
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1164, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1164, __pyx_L1_error) }
-      __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1164, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_v_tr = __pyx_t_1;
-      __pyx_t_1 = 0;
-
+1165:                     fr = classifyWithMissingData(observations, tree.falseBranch)
-
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1165, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1165, __pyx_L1_error) }
-      __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1165, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_v_fr = __pyx_t_4;
-      __pyx_t_4 = 0;
-
+1166:                     tcount = sum(tr.values())
-
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tr, __pyx_n_s_values); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1166, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_5 = NULL;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
-        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
-        if (likely(__pyx_t_5)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-          __Pyx_INCREF(__pyx_t_5);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_1, function);
-        }
-      }
-      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
-      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1166, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_t_1 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1166, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_v_tcount = __pyx_t_1;
-      __pyx_t_1 = 0;
-
+1167:                     fcount = sum(fr.values())
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_fr, __pyx_n_s_values); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1167, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_5 = NULL;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
-        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_4);
-        if (likely(__pyx_t_5)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-          __Pyx_INCREF(__pyx_t_5);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_4, function);
-        }
-      }
-      __pyx_t_1 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
-      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1167, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __pyx_t_4 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1167, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_v_fcount = __pyx_t_4;
-      __pyx_t_4 = 0;
-
+1168:                     tw = float(tcount) / (tcount + fcount)
-
      __pyx_t_4 = __Pyx_PyNumber_Float(__pyx_v_tcount); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1168, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_1 = PyNumber_Add(__pyx_v_tcount, __pyx_v_fcount); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1168, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_5 = __Pyx_PyNumber_Divide(__pyx_t_4, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1168, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_v_tw = __pyx_t_5;
-      __pyx_t_5 = 0;
-
+1169:                     fw = float(fcount) / (tcount + fcount)
-
      __pyx_t_5 = __Pyx_PyNumber_Float(__pyx_v_fcount); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1169, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __pyx_t_1 = PyNumber_Add(__pyx_v_tcount, __pyx_v_fcount); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1169, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_4 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1169, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_v_fw = __pyx_t_4;
-      __pyx_t_4 = 0;
-
 1170:
-
 1171:                     # Problem description: http://blog.ludovf.net/python-collections-defaultdict/
-
+1172:                     result = defaultdict(int)
-
      __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_defaultdict); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1172, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_5 = NULL;
-      if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_1))) {
-        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
-        if (likely(__pyx_t_5)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-          __Pyx_INCREF(__pyx_t_5);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_1, function);
-        }
-      }
-      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_1, __pyx_t_5, ((PyObject *)(&PyInt_Type))) : __Pyx_PyObject_CallOneArg(__pyx_t_1, ((PyObject *)(&PyInt_Type)));
-      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1172, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      __pyx_v_result = __pyx_t_4;
-      __pyx_t_4 = 0;
-
+1173:                     for k, v in tr.items():
-
      __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tr, __pyx_n_s_items); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __pyx_t_5 = NULL;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_1))) {
-        __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_1);
-        if (likely(__pyx_t_5)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_1);
-          __Pyx_INCREF(__pyx_t_5);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_1, function);
-        }
-      }
-      __pyx_t_4 = (__pyx_t_5) ? __Pyx_PyObject_CallOneArg(__pyx_t_1, __pyx_t_5) : __Pyx_PyObject_CallNoArg(__pyx_t_1);
-      __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-      if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
-        __pyx_t_1 = __pyx_t_4; __Pyx_INCREF(__pyx_t_1); __pyx_t_6 = 0;
-        __pyx_t_7 = NULL;
-      } else {
-        __pyx_t_6 = -1; __pyx_t_1 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1173, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        __pyx_t_7 = Py_TYPE(__pyx_t_1)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1173, __pyx_L1_error)
-      }
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      for (;;) {
-        if (likely(!__pyx_t_7)) {
-          if (likely(PyList_CheckExact(__pyx_t_1))) {
-            if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_1)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_4 = PyList_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1173, __pyx_L1_error)
-            #else
-            __pyx_t_4 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_4);
-            #endif
-          } else {
-            if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_1)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_1, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1173, __pyx_L1_error)
-            #else
-            __pyx_t_4 = PySequence_ITEM(__pyx_t_1, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1173, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_4);
-            #endif
-          }
-        } else {
-          __pyx_t_4 = __pyx_t_7(__pyx_t_1);
-          if (unlikely(!__pyx_t_4)) {
-            PyObject* exc_type = PyErr_Occurred();
-            if (exc_type) {
-              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-              else __PYX_ERR(0, 1173, __pyx_L1_error)
-            }
-            break;
-          }
-          __Pyx_GOTREF(__pyx_t_4);
-        }
-        if ((likely(PyTuple_CheckExact(__pyx_t_4))) || (PyList_CheckExact(__pyx_t_4))) {
-          PyObject* sequence = __pyx_t_4;
-          Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-          if (unlikely(size != 2)) {
-            if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-            else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-            __PYX_ERR(0, 1173, __pyx_L1_error)
-          }
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          if (likely(PyTuple_CheckExact(sequence))) {
-            __pyx_t_5 = PyTuple_GET_ITEM(sequence, 0); 
-            __pyx_t_8 = PyTuple_GET_ITEM(sequence, 1); 
-          } else {
-            __pyx_t_5 = PyList_GET_ITEM(sequence, 0); 
-            __pyx_t_8 = PyList_GET_ITEM(sequence, 1); 
-          }
-          __Pyx_INCREF(__pyx_t_5);
-          __Pyx_INCREF(__pyx_t_8);
-          #else
-          __pyx_t_5 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1173, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_5);
-          __pyx_t_8 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1173, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_8);
-          #endif
-          __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        } else {
-          Py_ssize_t index = -1;
-          __pyx_t_9 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1173, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-          __pyx_t_10 = Py_TYPE(__pyx_t_9)->tp_iternext;
-          index = 0; __pyx_t_5 = __pyx_t_10(__pyx_t_9); if (unlikely(!__pyx_t_5)) goto __pyx_L7_unpacking_failed;
-          __Pyx_GOTREF(__pyx_t_5);
-          index = 1; __pyx_t_8 = __pyx_t_10(__pyx_t_9); if (unlikely(!__pyx_t_8)) goto __pyx_L7_unpacking_failed;
-          __Pyx_GOTREF(__pyx_t_8);
-          if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_9), 2) < 0) __PYX_ERR(0, 1173, __pyx_L1_error)
-          __pyx_t_10 = NULL;
-          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-          goto __pyx_L8_unpacking_done;
-          __pyx_L7_unpacking_failed:;
-          __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-          __pyx_t_10 = NULL;
-          if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-          __PYX_ERR(0, 1173, __pyx_L1_error)
-          __pyx_L8_unpacking_done:;
-        }
-        __Pyx_XDECREF_SET(__pyx_v_k, __pyx_t_5);
-        __pyx_t_5 = 0;
-        __Pyx_DECREF_SET(__pyx_v_v, __pyx_t_8);
-        __pyx_t_8 = 0;
-/* … */
-      }
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+1174:                         result[k] += v * tw
-
        __Pyx_INCREF(__pyx_v_k);
-        __pyx_t_4 = __pyx_v_k;
-        __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_result, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1174, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __pyx_t_5 = PyNumber_Multiply(__pyx_v_v, __pyx_v_tw); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1174, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __pyx_t_9 = PyNumber_InPlaceAdd(__pyx_t_8, __pyx_t_5); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1174, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        if (unlikely(PyObject_SetItem(__pyx_v_result, __pyx_t_4, __pyx_t_9) < 0)) __PYX_ERR(0, 1174, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+1175:                     for k, v in fr.items():
-
      __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_fr, __pyx_n_s_items); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1175, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_9 = NULL;
-      if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_4))) {
-        __pyx_t_9 = PyMethod_GET_SELF(__pyx_t_4);
-        if (likely(__pyx_t_9)) {
-          PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_4);
-          __Pyx_INCREF(__pyx_t_9);
-          __Pyx_INCREF(function);
-          __Pyx_DECREF_SET(__pyx_t_4, function);
-        }
-      }
-      __pyx_t_1 = (__pyx_t_9) ? __Pyx_PyObject_CallOneArg(__pyx_t_4, __pyx_t_9) : __Pyx_PyObject_CallNoArg(__pyx_t_4);
-      __Pyx_XDECREF(__pyx_t_9); __pyx_t_9 = 0;
-      if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
-        __pyx_t_4 = __pyx_t_1; __Pyx_INCREF(__pyx_t_4); __pyx_t_6 = 0;
-        __pyx_t_7 = NULL;
-      } else {
-        __pyx_t_6 = -1; __pyx_t_4 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1175, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __pyx_t_7 = Py_TYPE(__pyx_t_4)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1175, __pyx_L1_error)
-      }
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      for (;;) {
-        if (likely(!__pyx_t_7)) {
-          if (likely(PyList_CheckExact(__pyx_t_4))) {
-            if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_4)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_1 = PyList_GET_ITEM(__pyx_t_4, __pyx_t_6); __Pyx_INCREF(__pyx_t_1); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1175, __pyx_L1_error)
-            #else
-            __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_1);
-            #endif
-          } else {
-            if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_4)) break;
-            #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-            __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_4, __pyx_t_6); __Pyx_INCREF(__pyx_t_1); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1175, __pyx_L1_error)
-            #else
-            __pyx_t_1 = PySequence_ITEM(__pyx_t_4, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1175, __pyx_L1_error)
-            __Pyx_GOTREF(__pyx_t_1);
-            #endif
-          }
-        } else {
-          __pyx_t_1 = __pyx_t_7(__pyx_t_4);
-          if (unlikely(!__pyx_t_1)) {
-            PyObject* exc_type = PyErr_Occurred();
-            if (exc_type) {
-              if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-              else __PYX_ERR(0, 1175, __pyx_L1_error)
-            }
-            break;
-          }
-          __Pyx_GOTREF(__pyx_t_1);
-        }
-        if ((likely(PyTuple_CheckExact(__pyx_t_1))) || (PyList_CheckExact(__pyx_t_1))) {
-          PyObject* sequence = __pyx_t_1;
-          Py_ssize_t size = __Pyx_PySequence_SIZE(sequence);
-          if (unlikely(size != 2)) {
-            if (size > 2) __Pyx_RaiseTooManyValuesError(2);
-            else if (size >= 0) __Pyx_RaiseNeedMoreValuesError(size);
-            __PYX_ERR(0, 1175, __pyx_L1_error)
-          }
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          if (likely(PyTuple_CheckExact(sequence))) {
-            __pyx_t_9 = PyTuple_GET_ITEM(sequence, 0); 
-            __pyx_t_5 = PyTuple_GET_ITEM(sequence, 1); 
-          } else {
-            __pyx_t_9 = PyList_GET_ITEM(sequence, 0); 
-            __pyx_t_5 = PyList_GET_ITEM(sequence, 1); 
-          }
-          __Pyx_INCREF(__pyx_t_9);
-          __Pyx_INCREF(__pyx_t_5);
-          #else
-          __pyx_t_9 = PySequence_ITEM(sequence, 0); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1175, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_9);
-          __pyx_t_5 = PySequence_ITEM(sequence, 1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1175, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_5);
-          #endif
-          __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-        } else {
-          Py_ssize_t index = -1;
-          __pyx_t_8 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1175, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_8);
-          __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-          __pyx_t_10 = Py_TYPE(__pyx_t_8)->tp_iternext;
-          index = 0; __pyx_t_9 = __pyx_t_10(__pyx_t_8); if (unlikely(!__pyx_t_9)) goto __pyx_L11_unpacking_failed;
-          __Pyx_GOTREF(__pyx_t_9);
-          index = 1; __pyx_t_5 = __pyx_t_10(__pyx_t_8); if (unlikely(!__pyx_t_5)) goto __pyx_L11_unpacking_failed;
-          __Pyx_GOTREF(__pyx_t_5);
-          if (__Pyx_IternextUnpackEndCheck(__pyx_t_10(__pyx_t_8), 2) < 0) __PYX_ERR(0, 1175, __pyx_L1_error)
-          __pyx_t_10 = NULL;
-          __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-          goto __pyx_L12_unpacking_done;
-          __pyx_L11_unpacking_failed:;
-          __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-          __pyx_t_10 = NULL;
-          if (__Pyx_IterFinish() == 0) __Pyx_RaiseNeedMoreValuesError(index);
-          __PYX_ERR(0, 1175, __pyx_L1_error)
-          __pyx_L12_unpacking_done:;
-        }
-        __Pyx_XDECREF_SET(__pyx_v_k, __pyx_t_9);
-        __pyx_t_9 = 0;
-        __Pyx_DECREF_SET(__pyx_v_v, __pyx_t_5);
-        __pyx_t_5 = 0;
-/* … */
-      }
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+1176:                         result[k] += v * fw
-
        __Pyx_INCREF(__pyx_v_k);
-        __pyx_t_1 = __pyx_v_k;
-        __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_result, __pyx_t_1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1176, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        __pyx_t_9 = PyNumber_Multiply(__pyx_v_v, __pyx_v_fw); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1176, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_9);
-        __pyx_t_8 = PyNumber_InPlaceAdd(__pyx_t_5, __pyx_t_9); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1176, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_8);
-        __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-        __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-        if (unlikely(PyObject_SetItem(__pyx_v_result, __pyx_t_1, __pyx_t_8) < 0)) __PYX_ERR(0, 1176, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+1177:                     return dict(result)
-
      __Pyx_XDECREF(__pyx_r);
-      __pyx_t_4 = __Pyx_PyObject_CallOneArg(((PyObject *)(&PyDict_Type)), __pyx_v_result); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1177, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_r = __pyx_t_4;
-      __pyx_t_4 = 0;
-      goto __pyx_L0;
-
 1178:                 else:
-
+1179:                     branch = None
-
    /*else*/ {
-      __Pyx_INCREF(Py_None);
-      __pyx_v_branch = Py_None;
-
+1180:                     if isinstance(v, int) or isinstance(v, float):
-
      __pyx_t_3 = PyInt_Check(__pyx_v_v); 
-      __pyx_t_11 = (__pyx_t_3 != 0);
-      if (!__pyx_t_11) {
-      } else {
-        __pyx_t_2 = __pyx_t_11;
-        goto __pyx_L14_bool_binop_done;
-      }
-      __pyx_t_11 = PyFloat_Check(__pyx_v_v); 
-      __pyx_t_3 = (__pyx_t_11 != 0);
-      __pyx_t_2 = __pyx_t_3;
-      __pyx_L14_bool_binop_done:;
-      if (__pyx_t_2) {
-/* … */
-        goto __pyx_L13;
-      }
-
+1181:                         if v >= tree.value:
-
        __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1181, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        __pyx_t_1 = PyObject_RichCompare(__pyx_v_v, __pyx_t_4, Py_GE); __Pyx_XGOTREF(__pyx_t_1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1181, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_1); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1181, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-        if (__pyx_t_2) {
-/* … */
-          goto __pyx_L16;
-        }
-
+1182:                             branch = tree.trueBranch
-
          __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1182, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_1);
-          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_1);
-          __pyx_t_1 = 0;
-
 1183:                         else:
-
+1184:                             branch = tree.falseBranch
-
        /*else*/ {
-          __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1184, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_1);
-          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_1);
-          __pyx_t_1 = 0;
-        }
-        __pyx_L16:;
-
 1185:                     else:
-
+1186:                         if v == tree.value:
-
      /*else*/ {
-        __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_value); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1186, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        __pyx_t_4 = PyObject_RichCompare(__pyx_v_v, __pyx_t_1, Py_EQ); __Pyx_XGOTREF(__pyx_t_4); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1186, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-        __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1186, __pyx_L1_error)
-        __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-        if (__pyx_t_2) {
-/* … */
-          goto __pyx_L17;
-        }
-
+1187:                             branch = tree.trueBranch
-
          __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_trueBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1187, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
-          __pyx_t_4 = 0;
-
 1188:                         else:
-
+1189:                             branch = tree.falseBranch
-
        /*else*/ {
-          __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_falseBranch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1189, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_4);
-          __Pyx_DECREF_SET(__pyx_v_branch, __pyx_t_4);
-          __pyx_t_4 = 0;
-        }
-        __pyx_L17:;
-      }
-      __pyx_L13:;
-    }
-
+1190:                 return classifyWithMissingData(observations, branch)
-
    __Pyx_XDECREF(__pyx_r);
-    if (unlikely(!__pyx_cur_scope->__pyx_v_classifyWithMissingData)) { __Pyx_RaiseClosureNameError("classifyWithMissingData"); __PYX_ERR(0, 1190, __pyx_L1_error) }
-    __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_v_branch); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1190, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_r = __pyx_t_4;
-    __pyx_t_4 = 0;
-    goto __pyx_L0;
-  }
-
 1191:
-
 1192:         # function body
-
+1193:         if dataMissing:
-
  __pyx_t_2 = __Pyx_PyObject_IsTrue(__pyx_v_dataMissing); if (unlikely(__pyx_t_2 < 0)) __PYX_ERR(0, 1193, __pyx_L1_error)
-  if (__pyx_t_2) {
-/* … */
-  }
-
+1194:             return classifyWithMissingData(observations, tree)
-
    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_2classifyWithMissingData(__pyx_cur_scope->__pyx_v_classifyWithMissingData, __pyx_v_observations, __pyx_v_tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1194, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_r = __pyx_t_1;
-    __pyx_t_1 = 0;
-    goto __pyx_L0;
-
 1195:         else:
-
+1196:             return classifyWithoutMissingData(observations, tree)
-
  /*else*/ {
-    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_1 = __pyx_pf_8causalml_9inference_4tree_6uplift_20UpliftTreeClassifier_8classify_classifyWithoutMissingData(__pyx_cur_scope->__pyx_v_classifyWithoutMissingData, __pyx_v_observations, __pyx_v_tree); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1196, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_r = __pyx_t_1;
-    __pyx_t_1 = 0;
-    goto __pyx_L0;
-  }
-
 1197:
-
 1198:
-
 1199: # Uplift Random Forests
-
+1200: class UpliftRandomForestClassifier:
-
  __pyx_t_1 = __Pyx_Py3MetaclassPrepare((PyObject *) NULL, __pyx_empty_tuple, __pyx_n_s_UpliftRandomForestClassifier, __pyx_n_s_UpliftRandomForestClassifier, (PyObject *) NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_kp_s_Uplift_Random_Forest_for_Classi); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1200, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  __pyx_t_2 = __Pyx_Py3ClassCreate(((PyObject*)&__Pyx_DefaultClassType), __pyx_n_s_UpliftRandomForestClassifier, __pyx_empty_tuple, __pyx_t_1, NULL, 0, 1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1200, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_d, __pyx_n_s_UpliftRandomForestClassifier, __pyx_t_2) < 0) __PYX_ERR(0, 1200, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 1201:     """ Uplift Random Forest for Classification Task.
-
 1202:
-
 1203:     Parameters
-
 1204:     ----------
-
 1205:     n_estimators : integer, optional (default=10)
-
 1206:         The number of trees in the uplift random forest.
-
 1207:
-
 1208:     evaluationFunction : string
-
 1209:         Choose from one of the models: 'KL', 'ED', 'Chi', 'CTS', 'DDP'.
-
 1210:
-
 1211:     max_features: int, optional (default=10)
-
 1212:         The number of features to consider when looking for the best split.
-
 1213:
-
 1214:     random_state: int, RandomState instance or None (default=None)
-
 1215:         A random seed or `np.random.RandomState` to control randomness in building the trees and forest.
-
 1216:
-
 1217:     max_depth: int, optional (default=5)
-
 1218:         The maximum depth of the tree.
-
 1219:
-
 1220:     min_samples_leaf: int, optional (default=100)
-
 1221:         The minimum number of samples required to be split at a leaf node.
-
 1222:
-
 1223:     min_samples_treatment: int, optional (default=10)
-
 1224:         The minimum number of samples required of the experiment group to be split at a leaf node.
-
 1225:
-
 1226:     n_reg: int, optional (default=10)
-
 1227:         The regularization parameter defined in Rzepakowski et al. 2012, the
-
 1228:         weight (in terms of sample size) of the parent node influence on the
-
 1229:         child node, only effective for 'KL', 'ED', 'Chi', 'CTS' methods.
-
 1230:
-
 1231:     control_name: string
-
 1232:         The name of the control group (other experiment groups will be regarded as treatment groups)
-
 1233:
-
 1234:     normalization: boolean, optional (default=True)
-
 1235:         The normalization factor defined in Rzepakowski et al. 2012,
-
 1236:         correcting for tests with large number of splits and imbalanced
-
 1237:         treatment and control splits
-
 1238:
-
 1239:     n_jobs: int, optional (default=-1)
-
 1240:         The parallelization parameter to define how many parallel jobs need to be created.
-
 1241:         This is passed on to joblib library for parallelizing uplift-tree creation and prediction.
-
 1242:
-
 1243:     joblib_prefer: str, optional (default="threads")
-
 1244:         The preferred backend for joblib (passed as `prefer` to joblib.Parallel). See the joblib
-
 1245:         documentation for valid values.
-
 1246:
-
 1247:     Outputs
-
 1248:     ----------
-
 1249:     df_res: pandas dataframe
-
 1250:         A user-level results dataframe containing the estimated individual treatment effect.
-
 1251:     """
-
+1252:     def __init__(self,
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__[] = "\n        Initialize the UpliftRandomForestClassifier class.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__ = {"__init__", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_control_name = 0;
-  PyObject *__pyx_v_n_estimators = 0;
-  PyObject *__pyx_v_max_features = 0;
-  PyObject *__pyx_v_random_state = 0;
-  PyObject *__pyx_v_max_depth = 0;
-  PyObject *__pyx_v_min_samples_leaf = 0;
-  PyObject *__pyx_v_min_samples_treatment = 0;
-  PyObject *__pyx_v_n_reg = 0;
-  PyObject *__pyx_v_evaluationFunction = 0;
-  PyObject *__pyx_v_normalization = 0;
-  PyObject *__pyx_v_n_jobs = 0;
-  PyObject *__pyx_v_joblib_prefer = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__ (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_control_name,&__pyx_n_s_n_estimators,&__pyx_n_s_max_features,&__pyx_n_s_random_state,&__pyx_n_s_max_depth,&__pyx_n_s_min_samples_leaf,&__pyx_n_s_min_samples_treatment,&__pyx_n_s_n_reg,&__pyx_n_s_evaluationFunction,&__pyx_n_s_normalization,&__pyx_n_s_n_jobs,&__pyx_n_s_joblib_prefer,0};
-    PyObject* values[13] = {0,0,0,0,0,0,0,0,0,0,0,0,0};
-    values[2] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[3] = ((PyObject *)((PyObject *)__pyx_int_10));
-/* … */
-  /* function exit code */
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_control_name, PyObject *__pyx_v_n_estimators, PyObject *__pyx_v_max_features, PyObject *__pyx_v_random_state, PyObject *__pyx_v_max_depth, PyObject *__pyx_v_min_samples_leaf, PyObject *__pyx_v_min_samples_treatment, PyObject *__pyx_v_n_reg, PyObject *__pyx_v_evaluationFunction, PyObject *__pyx_v_normalization, PyObject *__pyx_v_n_jobs, PyObject *__pyx_v_joblib_prefer) {
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("__init__", 0);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__64 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_control_name, __pyx_n_s_n_estimators, __pyx_n_s_max_features, __pyx_n_s_random_state, __pyx_n_s_max_depth, __pyx_n_s_min_samples_leaf, __pyx_n_s_min_samples_treatment, __pyx_n_s_n_reg, __pyx_n_s_evaluationFunction, __pyx_n_s_normalization, __pyx_n_s_n_jobs, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_tuple__64)) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__64);
-  __Pyx_GIVEREF(__pyx_tuple__64);
-  __pyx_codeobj__65 = (PyObject*)__Pyx_PyCode_New(13, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__64, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_init, 1252, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__65)) __PYX_ERR(0, 1252, __pyx_L1_error)
-/* … */
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_joblib_prefer, __pyx_n_u_str) < 0) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_1__init__, 0, __pyx_n_s_UpliftRandomForestClassifier___i, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__65)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_5, __pyx_tuple__66);
-  __Pyx_CyFunction_SetAnnotationsDict(__pyx_t_5, __pyx_t_2);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_init, __pyx_t_5) < 0) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_tuple__66 = PyTuple_Pack(11, ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_10), ((PyObject *)Py_None), ((PyObject *)__pyx_int_5), ((PyObject *)__pyx_int_100), ((PyObject *)__pyx_int_10), ((PyObject *)__pyx_int_10), ((PyObject*)__pyx_n_s_KL), ((PyObject *)Py_True), ((PyObject *)__pyx_int_neg_1), ((PyObject*)__pyx_n_s_threads)); if (unlikely(!__pyx_tuple__66)) __PYX_ERR(0, 1252, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__66);
-  __Pyx_GIVEREF(__pyx_tuple__66);
-
 1253:                  control_name,
-
 1254:                  n_estimators=10,
-
 1255:                  max_features=10,
-
+1256:                  random_state=None,
-
    values[4] = ((PyObject *)((PyObject *)Py_None));
-    values[5] = ((PyObject *)((PyObject *)__pyx_int_5));
-    values[6] = ((PyObject *)((PyObject *)__pyx_int_100));
-    values[7] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[8] = ((PyObject *)((PyObject *)__pyx_int_10));
-    values[9] = ((PyObject *)((PyObject*)__pyx_n_s_KL));
-
 1257:                  max_depth=5,
-
 1258:                  min_samples_leaf=100,
-
 1259:                  min_samples_treatment=10,
-
 1260:                  n_reg=10,
-
 1261:                  evaluationFunction='KL',
-
+1262:                  normalization=True,
-
    values[10] = ((PyObject *)((PyObject *)Py_True));
-    values[11] = ((PyObject *)((PyObject *)__pyx_int_neg_1));
-    values[12] = ((PyObject*)((PyObject*)__pyx_n_s_threads));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
-        CYTHON_FALLTHROUGH;
-        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
-        CYTHON_FALLTHROUGH;
-        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
-        CYTHON_FALLTHROUGH;
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_control_name)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 13, 1); __PYX_ERR(0, 1252, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_estimators);
-          if (value) { values[2] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_features);
-          if (value) { values[3] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  4:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_random_state);
-          if (value) { values[4] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  5:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_max_depth);
-          if (value) { values[5] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  6:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_leaf);
-          if (value) { values[6] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  7:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_min_samples_treatment);
-          if (value) { values[7] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  8:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_reg);
-          if (value) { values[8] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case  9:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_evaluationFunction);
-          if (value) { values[9] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 10:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_normalization);
-          if (value) { values[10] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 11:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_n_jobs);
-          if (value) { values[11] = value; kw_args--; }
-        }
-        CYTHON_FALLTHROUGH;
-        case 12:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_joblib_prefer);
-          if (value) { values[12] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "__init__") < 0)) __PYX_ERR(0, 1252, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case 13: values[12] = PyTuple_GET_ITEM(__pyx_args, 12);
-        CYTHON_FALLTHROUGH;
-        case 12: values[11] = PyTuple_GET_ITEM(__pyx_args, 11);
-        CYTHON_FALLTHROUGH;
-        case 11: values[10] = PyTuple_GET_ITEM(__pyx_args, 10);
-        CYTHON_FALLTHROUGH;
-        case 10: values[9] = PyTuple_GET_ITEM(__pyx_args, 9);
-        CYTHON_FALLTHROUGH;
-        case  9: values[8] = PyTuple_GET_ITEM(__pyx_args, 8);
-        CYTHON_FALLTHROUGH;
-        case  8: values[7] = PyTuple_GET_ITEM(__pyx_args, 7);
-        CYTHON_FALLTHROUGH;
-        case  7: values[6] = PyTuple_GET_ITEM(__pyx_args, 6);
-        CYTHON_FALLTHROUGH;
-        case  6: values[5] = PyTuple_GET_ITEM(__pyx_args, 5);
-        CYTHON_FALLTHROUGH;
-        case  5: values[4] = PyTuple_GET_ITEM(__pyx_args, 4);
-        CYTHON_FALLTHROUGH;
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_control_name = values[1];
-    __pyx_v_n_estimators = values[2];
-    __pyx_v_max_features = values[3];
-    __pyx_v_random_state = values[4];
-    __pyx_v_max_depth = values[5];
-    __pyx_v_min_samples_leaf = values[6];
-    __pyx_v_min_samples_treatment = values[7];
-    __pyx_v_n_reg = values[8];
-    __pyx_v_evaluationFunction = values[9];
-    __pyx_v_normalization = values[10];
-    __pyx_v_n_jobs = values[11];
-    __pyx_v_joblib_prefer = ((PyObject*)values[12]);
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("__init__", 0, 2, 13, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1252, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.__init__", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  if (unlikely(!__Pyx_ArgTypeTest(((PyObject *)__pyx_v_joblib_prefer), (&PyString_Type), 1, "joblib_prefer", 1))) __PYX_ERR(0, 1264, __pyx_L1_error)
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier___init__(__pyx_self, __pyx_v_self, __pyx_v_control_name, __pyx_v_n_estimators, __pyx_v_max_features, __pyx_v_random_state, __pyx_v_max_depth, __pyx_v_min_samples_leaf, __pyx_v_min_samples_treatment, __pyx_v_n_reg, __pyx_v_evaluationFunction, __pyx_v_normalization, __pyx_v_n_jobs, __pyx_v_joblib_prefer);
-
 1263:                  n_jobs=-1,
-
 1264:                  joblib_prefer: str = "threads"):
-
 1265:
-
 1266:         """
-
 1267:         Initialize the UpliftRandomForestClassifier class.
-
 1268:         """
-
+1269:         self.n_estimators = n_estimators
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_estimators, __pyx_v_n_estimators) < 0) __PYX_ERR(0, 1269, __pyx_L1_error)
-
+1270:         self.max_features = max_features
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_features, __pyx_v_max_features) < 0) __PYX_ERR(0, 1270, __pyx_L1_error)
-
+1271:         self.random_state = random_state
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_random_state, __pyx_v_random_state) < 0) __PYX_ERR(0, 1271, __pyx_L1_error)
-
+1272:         self.max_depth = max_depth
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_max_depth, __pyx_v_max_depth) < 0) __PYX_ERR(0, 1272, __pyx_L1_error)
-
+1273:         self.min_samples_leaf = min_samples_leaf
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_leaf, __pyx_v_min_samples_leaf) < 0) __PYX_ERR(0, 1273, __pyx_L1_error)
-
+1274:         self.min_samples_treatment = min_samples_treatment
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_min_samples_treatment, __pyx_v_min_samples_treatment) < 0) __PYX_ERR(0, 1274, __pyx_L1_error)
-
+1275:         self.n_reg = n_reg
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_reg, __pyx_v_n_reg) < 0) __PYX_ERR(0, 1275, __pyx_L1_error)
-
+1276:         self.evaluationFunction = evaluationFunction
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_evaluationFunction, __pyx_v_evaluationFunction) < 0) __PYX_ERR(0, 1276, __pyx_L1_error)
-
+1277:         self.control_name = control_name
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 1277, __pyx_L1_error)
-
+1278:         self.normalization = normalization
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_normalization, __pyx_v_normalization) < 0) __PYX_ERR(0, 1278, __pyx_L1_error)
-
+1279:         self.n_jobs = n_jobs
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs, __pyx_v_n_jobs) < 0) __PYX_ERR(0, 1279, __pyx_L1_error)
-
+1280:         self.joblib_prefer = joblib_prefer
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_joblib_prefer, __pyx_v_joblib_prefer) < 0) __PYX_ERR(0, 1280, __pyx_L1_error)
-
 1281:
-
+1282:         assert control_name is not None and isinstance(control_name, str), \
-
  #ifndef CYTHON_WITHOUT_ASSERTIONS
-  if (unlikely(!Py_OptimizeFlag)) {
-    __pyx_t_2 = (__pyx_v_control_name != Py_None);
-    __pyx_t_3 = (__pyx_t_2 != 0);
-    if (__pyx_t_3) {
-    } else {
-      __pyx_t_1 = __pyx_t_3;
-      goto __pyx_L3_bool_binop_done;
-    }
-    __pyx_t_3 = PyString_Check(__pyx_v_control_name); 
-    __pyx_t_2 = (__pyx_t_3 != 0);
-    __pyx_t_1 = __pyx_t_2;
-    __pyx_L3_bool_binop_done:;
-    if (unlikely(!__pyx_t_1)) {
-
+1283:             f"control_group should be string but {control_name} is passed"
-
      __pyx_t_4 = PyTuple_New(3); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1283, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_4);
-      __pyx_t_5 = 0;
-      __pyx_t_6 = 127;
-      __Pyx_INCREF(__pyx_kp_u_control_group_should_be_string_b);
-      __pyx_t_5 += 35;
-      __Pyx_GIVEREF(__pyx_kp_u_control_group_should_be_string_b);
-      PyTuple_SET_ITEM(__pyx_t_4, 0, __pyx_kp_u_control_group_should_be_string_b);
-      __pyx_t_7 = __Pyx_PyObject_FormatSimple(__pyx_v_control_name, __pyx_empty_unicode); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1283, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __pyx_t_6 = (__Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) > __pyx_t_6) ? __Pyx_PyUnicode_MAX_CHAR_VALUE(__pyx_t_7) : __pyx_t_6;
-      __pyx_t_5 += __Pyx_PyUnicode_GET_LENGTH(__pyx_t_7);
-      __Pyx_GIVEREF(__pyx_t_7);
-      PyTuple_SET_ITEM(__pyx_t_4, 1, __pyx_t_7);
-      __pyx_t_7 = 0;
-      __Pyx_INCREF(__pyx_kp_u_is_passed);
-      __pyx_t_5 += 10;
-      __Pyx_GIVEREF(__pyx_kp_u_is_passed);
-      PyTuple_SET_ITEM(__pyx_t_4, 2, __pyx_kp_u_is_passed);
-      __pyx_t_7 = __Pyx_PyUnicode_Join(__pyx_t_4, 3, __pyx_t_5, __pyx_t_6); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1283, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-      PyErr_SetObject(PyExc_AssertionError, __pyx_t_7);
-      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-      __PYX_ERR(0, 1282, __pyx_L1_error)
-    }
-  }
-  #endif
-
+1284:         self.control_name = control_name
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_control_name, __pyx_v_control_name) < 0) __PYX_ERR(0, 1284, __pyx_L1_error)
-
+1285:         self.classes_ = [control_name]
-
  __pyx_t_7 = PyList_New(1); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1285, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_INCREF(__pyx_v_control_name);
-  __Pyx_GIVEREF(__pyx_v_control_name);
-  PyList_SET_ITEM(__pyx_t_7, 0, __pyx_v_control_name);
-  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_classes, __pyx_t_7) < 0) __PYX_ERR(0, 1285, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1286:         self.n_class = 1
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_class, __pyx_int_1) < 0) __PYX_ERR(0, 1286, __pyx_L1_error)
-
 1287:
-
+1288:         if self.n_jobs == -1:
-
  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_4 = __Pyx_PyInt_EqObjC(__pyx_t_7, __pyx_int_neg_1, -1L, 0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1288, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_t_1 = __Pyx_PyObject_IsTrue(__pyx_t_4); if (unlikely(__pyx_t_1 < 0)) __PYX_ERR(0, 1288, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  if (__pyx_t_1) {
-/* … */
-  }
-
+1289:             self.n_jobs = mp.cpu_count()
-
    __Pyx_GetModuleGlobalName(__pyx_t_7, __pyx_n_s_mp); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_7, __pyx_n_s_cpu_count); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __pyx_t_7 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_8))) {
-      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_8);
-      if (likely(__pyx_t_7)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_8);
-        __Pyx_INCREF(__pyx_t_7);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_8, function);
-      }
-    }
-    __pyx_t_4 = (__pyx_t_7) ? __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_7) : __Pyx_PyObject_CallNoArg(__pyx_t_8);
-    __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-    if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1289, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (__Pyx_PyObject_SetAttrStr(__pyx_v_self, __pyx_n_s_n_jobs, __pyx_t_4) < 0) __PYX_ERR(0, 1289, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
 1290:
-
+1291:     def fit(self, X, treatment, y):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit[] = "\n        Fit the UpliftRandomForestClassifier.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        treatment : array-like, shape = [num_samples]\n            An array containing the treatment group for each unit.\n\n        y : array-like, shape = [num_samples]\n            An array containing the outcome of interest for each unit.\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit = {"fit", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 1); __PYX_ERR(0, 1291, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 2); __PYX_ERR(0, 1291, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, 3); __PYX_ERR(0, 1291, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "fit") < 0)) __PYX_ERR(0, 1291, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_treatment = values[2];
-    __pyx_v_y = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("fit", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1291, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value); /* proto */
-/* … */
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_2fit(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *__pyx_cur_scope;
-  PyObject *__pyx_v_random_state = NULL;
-  PyObject *__pyx_v_treatment_groups = NULL;
-  PyObject *__pyx_v_tr = NULL;
-  PyObject *__pyx_v_all_importances = NULL;
-  CYTHON_UNUSED PyObject *__pyx_v__ = NULL;
-  PyObject *__pyx_v_x = NULL;
-  PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator = 0;
-  PyObject *__pyx_v_tree = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("fit", 0);
-  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit, __pyx_empty_tuple, NULL);
-  if (unlikely(!__pyx_cur_scope)) {
-    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *)Py_None);
-    __Pyx_INCREF(Py_None);
-    __PYX_ERR(0, 1291, __pyx_L1_error)
-  } else {
-    __Pyx_GOTREF(__pyx_cur_scope);
-  }
-  __pyx_cur_scope->__pyx_v_self = __pyx_v_self;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_self);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_self);
-  __pyx_cur_scope->__pyx_v_X = __pyx_v_X;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_X);
-  __pyx_cur_scope->__pyx_v_treatment = __pyx_v_treatment;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_treatment);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_treatment);
-  __pyx_cur_scope->__pyx_v_y = __pyx_v_y;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_y);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_y);
-/* … */
-  /* function exit code */
-  __pyx_r = Py_None; __Pyx_INCREF(Py_None);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_XDECREF(__pyx_t_9);
-  __Pyx_XDECREF(__pyx_t_10);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_random_state);
-  __Pyx_XDECREF(__pyx_v_treatment_groups);
-  __Pyx_XDECREF(__pyx_v_tr);
-  __Pyx_XDECREF(__pyx_v_all_importances);
-  __Pyx_XDECREF(__pyx_v__);
-  __Pyx_XDECREF(__pyx_v_x);
-  __Pyx_XDECREF(__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator);
-  __Pyx_XDECREF(__pyx_v_tree);
-  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__67 = PyTuple_Pack(13, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_random_state, __pyx_n_s_treatment_groups, __pyx_n_s_tr, __pyx_n_s_all_importances, __pyx_n_s__34, __pyx_n_s_x, __pyx_n_s_genexpr, __pyx_n_s_genexpr, __pyx_n_s_tree); if (unlikely(!__pyx_tuple__67)) __PYX_ERR(0, 1291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__67);
-  __Pyx_GIVEREF(__pyx_tuple__67);
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit, 0, __pyx_n_s_UpliftRandomForestClassifier_fit_2, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__68)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1291, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_fit, __pyx_t_5) < 0) __PYX_ERR(0, 1291, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_codeobj__68 = (PyObject*)__Pyx_PyCode_New(4, 0, 13, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__67, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_fit, 1291, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__68)) __PYX_ERR(0, 1291, __pyx_L1_error)
-/* … */
-struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit {
-  PyObject_HEAD
-  PyObject *__pyx_v_X;
-  PyObject *__pyx_v_self;
-  PyObject *__pyx_v_treatment;
-  PyObject *__pyx_v_y;
-};
-
-
 1292:         """
-
 1293:         Fit the UpliftRandomForestClassifier.
-
 1294:
-
 1295:         Args
-
 1296:         ----
-
 1297:         X : ndarray, shape = [num_samples, num_features]
-
 1298:             An ndarray of the covariates used to train the uplift model.
-
 1299:
-
 1300:         treatment : array-like, shape = [num_samples]
-
 1301:             An array containing the treatment group for each unit.
-
 1302:
-
 1303:         y : array-like, shape = [num_samples]
-
 1304:             An array containing the outcome of interest for each unit.
-
 1305:         """
-
+1306:         random_state = check_random_state(self.random_state)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1306, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_random_state = __pyx_t_1;
-  __pyx_t_1 = 0;
-
 1307:
-
 1308:         # Create forest
-
+1309:         self.uplift_forest = [
-
  __pyx_t_1 = PyList_New(0); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1309, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-/* … */
-  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest, __pyx_t_1) < 0) __PYX_ERR(0, 1309, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+1310:             UpliftTreeClassifier(
-
    __Pyx_GetModuleGlobalName(__pyx_t_3, __pyx_n_s_UpliftTreeClassifier); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1310, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-/* … */
-    __pyx_t_7 = __Pyx_PyObject_Call(__pyx_t_3, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1310, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (unlikely(__Pyx_ListComp_Append(__pyx_t_1, (PyObject*)__pyx_t_7))) __PYX_ERR(0, 1309, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1311:                 max_features=self.max_features, max_depth=self.max_depth,
-
    __pyx_t_4 = __Pyx_PyDict_NewPresized(9); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_max_features); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_max_features, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_max_depth); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_max_depth, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1312:                 min_samples_leaf=self.min_samples_leaf,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_min_samples_leaf); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1312, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_min_samples_leaf, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1313:                 min_samples_treatment=self.min_samples_treatment,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_min_samples_treatment); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1313, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_min_samples_treatment, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1314:                 n_reg=self.n_reg,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_reg); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1314, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_n_reg, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1315:                 evaluationFunction=self.evaluationFunction,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_evaluationFunction); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1315, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_evaluationFunction, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1316:                 control_name=self.control_name,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1316, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_control_name, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1317:                 normalization=self.normalization,
-
    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_normalization); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1317, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_normalization, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1318:                 random_state=random_state.randint(MAX_INT))
-
    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_v_random_state, __pyx_n_s_randint); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1318, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __Pyx_GetModuleGlobalName(__pyx_t_9, __pyx_n_s_MAX_INT); if (unlikely(!__pyx_t_9)) __PYX_ERR(0, 1318, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_9);
-    __pyx_t_10 = NULL;
-    if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_8))) {
-      __pyx_t_10 = PyMethod_GET_SELF(__pyx_t_8);
-      if (likely(__pyx_t_10)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_8);
-        __Pyx_INCREF(__pyx_t_10);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_8, function);
-      }
-    }
-    __pyx_t_7 = (__pyx_t_10) ? __Pyx_PyObject_Call2Args(__pyx_t_8, __pyx_t_10, __pyx_t_9) : __Pyx_PyObject_CallOneArg(__pyx_t_8, __pyx_t_9);
-    __Pyx_XDECREF(__pyx_t_10); __pyx_t_10 = 0;
-    __Pyx_DECREF(__pyx_t_9); __pyx_t_9 = 0;
-    if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1318, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_random_state, __pyx_t_7) < 0) __PYX_ERR(0, 1311, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1319:             for _ in range(self.n_estimators)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_estimators); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1319, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_CallOneArg(__pyx_builtin_range, __pyx_t_2); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
-    __pyx_t_2 = __pyx_t_3; __Pyx_INCREF(__pyx_t_2); __pyx_t_5 = 0;
-    __pyx_t_6 = NULL;
-  } else {
-    __pyx_t_5 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1319, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_6 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1319, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_6)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1319, __pyx_L1_error)
-        #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        #endif
-      } else {
-        if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1319, __pyx_L1_error)
-        #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1319, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        #endif
-      }
-    } else {
-      __pyx_t_3 = __pyx_t_6(__pyx_t_2);
-      if (unlikely(!__pyx_t_3)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1319, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_3);
-    }
-    __Pyx_XDECREF_SET(__pyx_v__, __pyx_t_3);
-    __pyx_t_3 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 1320:         ]
-
 1321:
-
 1322:         # Get treatment group keys. self.classes_[0] is reserved for the control group.
-
+1323:         treatment_groups = sorted([x for x in list(set(treatment)) if x != self.control_name])
-
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_7 = PySet_New(__pyx_cur_scope->__pyx_v_treatment); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_4 = PySequence_List(__pyx_t_7); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_t_7 = __pyx_t_4; __Pyx_INCREF(__pyx_t_7); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  for (;;) {
-    if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_7)) break;
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_4 = PyList_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_4); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1323, __pyx_L1_error)
-    #else
-    __pyx_t_4 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    #endif
-    __Pyx_XDECREF_SET(__pyx_v_x, __pyx_t_4);
-    __pyx_t_4 = 0;
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1323, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_3 = PyObject_RichCompare(__pyx_v_x, __pyx_t_4, Py_NE); __Pyx_XGOTREF(__pyx_t_3); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1323, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_11 = __Pyx_PyObject_IsTrue(__pyx_t_3); if (unlikely(__pyx_t_11 < 0)) __PYX_ERR(0, 1323, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    if (__pyx_t_11) {
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_v_x))) __PYX_ERR(0, 1323, __pyx_L1_error)
-    }
-  }
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_t_1 = ((PyObject*)__pyx_t_2);
-  __pyx_t_2 = 0;
-  __pyx_t_12 = PyList_Sort(__pyx_t_1); if (unlikely(__pyx_t_12 == ((int)-1))) __PYX_ERR(0, 1323, __pyx_L1_error)
-  __pyx_v_treatment_groups = ((PyObject*)__pyx_t_1);
-  __pyx_t_1 = 0;
-
+1324:         self.classes_ = [self.control_name]
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = PyList_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1324, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_GIVEREF(__pyx_t_1);
-  PyList_SET_ITEM(__pyx_t_2, 0, __pyx_t_1);
-  __pyx_t_1 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes, __pyx_t_2) < 0) __PYX_ERR(0, 1324, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1325:         for tr in treatment_groups:
-
  if (unlikely(__pyx_v_treatment_groups == Py_None)) {
-    PyErr_SetString(PyExc_TypeError, "'NoneType' object is not iterable");
-    __PYX_ERR(0, 1325, __pyx_L1_error)
-  }
-  __pyx_t_2 = __pyx_v_treatment_groups; __Pyx_INCREF(__pyx_t_2); __pyx_t_5 = 0;
-  for (;;) {
-    if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_2)) break;
-    #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-    __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_5); __Pyx_INCREF(__pyx_t_1); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1325, __pyx_L1_error)
-    #else
-    __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1325, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    #endif
-    __Pyx_XDECREF_SET(__pyx_v_tr, __pyx_t_1);
-    __pyx_t_1 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1326:             self.classes_.append(tr)
-
    __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1326, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_12 = __Pyx_PyObject_Append(__pyx_t_1, __pyx_v_tr); if (unlikely(__pyx_t_12 == ((int)-1))) __PYX_ERR(0, 1326, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+1327:         self.n_class = len(self.classes_)
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1327, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = PyObject_Length(__pyx_t_2); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1327, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1327, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_class, __pyx_t_2) < 0) __PYX_ERR(0, 1327, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
 1328:
-
+1329:         self.uplift_forest = (
-
  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest, __pyx_t_2) < 0) __PYX_ERR(0, 1329, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1330:             Parallel(n_jobs=self.n_jobs, prefer=self.joblib_prefer)
-
  __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_7 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_7, __pyx_n_s_n_jobs, __pyx_t_3) < 0) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_7, __pyx_n_s_prefer, __pyx_t_3) < 0) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  __pyx_t_3 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_7); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1330, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
+1331:             (delayed(self.bootstrap)(X, treatment, y, tree) for tree in self.uplift_forest)
-
static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_genexpr(PyObject *__pyx_self) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *__pyx_cur_scope;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("genexpr", 0);
-  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr, __pyx_empty_tuple, NULL);
-  if (unlikely(!__pyx_cur_scope)) {
-    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr *)Py_None);
-    __Pyx_INCREF(Py_None);
-    __PYX_ERR(0, 1331, __pyx_L1_error)
-  } else {
-    __Pyx_GOTREF(__pyx_cur_scope);
-  }
-  __pyx_cur_scope->__pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *) __pyx_self;
-  __Pyx_INCREF(((PyObject *)__pyx_cur_scope->__pyx_outer_scope));
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope);
-  {
-    __pyx_CoroutineObject *gen = __Pyx_Generator_New((__pyx_coroutine_body_t) __pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator, NULL, (PyObject *) __pyx_cur_scope, __pyx_n_s_genexpr, __pyx_n_s_UpliftRandomForestClassifier_fit, __pyx_n_s_causalml_inference_tree_uplift); if (unlikely(!gen)) __PYX_ERR(0, 1331, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_cur_scope);
-    __Pyx_RefNannyFinishContext();
-    return (PyObject *) gen;
-  }
-
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.fit.genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_2generator(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value) /* generator body */
-{
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("genexpr", 0);
-  __pyx_L3_first_run:;
-  if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1331, __pyx_L1_error) }
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
-    __pyx_t_2 = __pyx_t_1; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1331, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1331, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1331, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      }
-    } else {
-      __pyx_t_1 = __pyx_t_4(__pyx_t_2);
-      if (unlikely(!__pyx_t_1)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1331, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_1);
-    }
-    __Pyx_XGOTREF(__pyx_cur_scope->__pyx_v_tree);
-    __Pyx_XDECREF_SET(__pyx_cur_scope->__pyx_v_tree, __pyx_t_1);
-    __Pyx_GIVEREF(__pyx_t_1);
-    __pyx_t_1 = 0;
-    __Pyx_GetModuleGlobalName(__pyx_t_6, __pyx_n_s_delayed); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1331, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1331, __pyx_L1_error) }
-    __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_bootstrap); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_8 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
-      __pyx_t_8 = PyMethod_GET_SELF(__pyx_t_6);
-      if (likely(__pyx_t_8)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-        __Pyx_INCREF(__pyx_t_8);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_6, function);
-      }
-    }
-    __pyx_t_5 = (__pyx_t_8) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_8, __pyx_t_7) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_7);
-    __Pyx_XDECREF(__pyx_t_8); __pyx_t_8 = 0;
-    __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1331, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X)) { __Pyx_RaiseClosureNameError("X"); __PYX_ERR(0, 1331, __pyx_L1_error) }
-    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment)) { __Pyx_RaiseClosureNameError("treatment"); __PYX_ERR(0, 1331, __pyx_L1_error) }
-    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y)) { __Pyx_RaiseClosureNameError("y"); __PYX_ERR(0, 1331, __pyx_L1_error) }
-    __pyx_t_6 = NULL;
-    __pyx_t_9 = 0;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-      __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_5);
-      if (likely(__pyx_t_6)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-        __Pyx_INCREF(__pyx_t_6);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_5, function);
-        __pyx_t_9 = 1;
-      }
-    }
-    #if CYTHON_FAST_PYCALL
-    if (PyFunction_Check(__pyx_t_5)) {
-      PyObject *__pyx_temp[5] = {__pyx_t_6, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y, __pyx_cur_scope->__pyx_v_tree};
-      __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_9, 4+__pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_1);
-    } else
-    #endif
-    #if CYTHON_FAST_PYCCALL
-    if (__Pyx_PyFastCFunction_Check(__pyx_t_5)) {
-      PyObject *__pyx_temp[5] = {__pyx_t_6, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y, __pyx_cur_scope->__pyx_v_tree};
-      __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_5, __pyx_temp+1-__pyx_t_9, 4+__pyx_t_9); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-      __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-      __Pyx_GOTREF(__pyx_t_1);
-    } else
-    #endif
-    {
-      __pyx_t_7 = PyTuple_New(4+__pyx_t_9); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_7);
-      if (__pyx_t_6) {
-        __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_7, 0, __pyx_t_6); __pyx_t_6 = NULL;
-      }
-      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
-      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
-      PyTuple_SET_ITEM(__pyx_t_7, 0+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X);
-      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
-      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
-      PyTuple_SET_ITEM(__pyx_t_7, 1+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_treatment);
-      __Pyx_INCREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
-      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
-      PyTuple_SET_ITEM(__pyx_t_7, 2+__pyx_t_9, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_y);
-      __Pyx_INCREF(__pyx_cur_scope->__pyx_v_tree);
-      __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_tree);
-      PyTuple_SET_ITEM(__pyx_t_7, 3+__pyx_t_9, __pyx_cur_scope->__pyx_v_tree);
-      __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_t_7, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1331, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_r = __pyx_t_1;
-    __pyx_t_1 = 0;
-    __Pyx_XGIVEREF(__pyx_t_2);
-    __pyx_cur_scope->__pyx_t_0 = __pyx_t_2;
-    __pyx_cur_scope->__pyx_t_1 = __pyx_t_3;
-    __pyx_cur_scope->__pyx_t_2 = __pyx_t_4;
-    __Pyx_XGIVEREF(__pyx_r);
-    __Pyx_RefNannyFinishContext();
-    __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
-    /* return from generator, yielding value */
-    __pyx_generator->resume_label = 1;
-    return __pyx_r;
-    __pyx_L6_resume_from_yield:;
-    __pyx_t_2 = __pyx_cur_scope->__pyx_t_0;
-    __pyx_cur_scope->__pyx_t_0 = 0;
-    __Pyx_XGOTREF(__pyx_t_2);
-    __pyx_t_3 = __pyx_cur_scope->__pyx_t_1;
-    __pyx_t_4 = __pyx_cur_scope->__pyx_t_2;
-    if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  CYTHON_MAYBE_UNUSED_VAR(__pyx_cur_scope);
-
-  /* function exit code */
-  PyErr_SetNone(PyExc_StopIteration);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_AddTraceback("genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_r); __pyx_r = 0;
-  #if !CYTHON_USE_EXC_INFO_STACK
-  __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
-  #endif
-  __pyx_generator->resume_label = -1;
-  __Pyx_Coroutine_clear((PyObject*)__pyx_generator);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_t_7 = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_3fit_genexpr(((PyObject*)__pyx_cur_scope)); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __pyx_t_1 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_3))) {
-    __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_3);
-    if (likely(__pyx_t_1)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_3);
-      __Pyx_INCREF(__pyx_t_1);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_3, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_3, __pyx_t_1, __pyx_t_7) : __Pyx_PyObject_CallOneArg(__pyx_t_3, __pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1331, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-/* … */
-struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_2_genexpr {
-  PyObject_HEAD
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_1_fit *__pyx_outer_scope;
-  PyObject *__pyx_v_tree;
-  PyObject *__pyx_t_0;
-  Py_ssize_t __pyx_t_1;
-  PyObject *(*__pyx_t_2)(PyObject *);
-};
-
-
 1332:         )
-
 1333:
-
+1334:         all_importances = [tree.feature_importances_ for tree in self.uplift_forest]
-
  __pyx_t_2 = PyList_New(0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1334, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (likely(PyList_CheckExact(__pyx_t_3)) || PyTuple_CheckExact(__pyx_t_3)) {
-    __pyx_t_7 = __pyx_t_3; __Pyx_INCREF(__pyx_t_7); __pyx_t_5 = 0;
-    __pyx_t_6 = NULL;
-  } else {
-    __pyx_t_5 = -1; __pyx_t_7 = PyObject_GetIter(__pyx_t_3); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1334, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_7);
-    __pyx_t_6 = Py_TYPE(__pyx_t_7)->tp_iternext; if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1334, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_6)) {
-      if (likely(PyList_CheckExact(__pyx_t_7))) {
-        if (__pyx_t_5 >= PyList_GET_SIZE(__pyx_t_7)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyList_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1334, __pyx_L1_error)
-        #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        #endif
-      } else {
-        if (__pyx_t_5 >= PyTuple_GET_SIZE(__pyx_t_7)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_3 = PyTuple_GET_ITEM(__pyx_t_7, __pyx_t_5); __Pyx_INCREF(__pyx_t_3); __pyx_t_5++; if (unlikely(0 < 0)) __PYX_ERR(0, 1334, __pyx_L1_error)
-        #else
-        __pyx_t_3 = PySequence_ITEM(__pyx_t_7, __pyx_t_5); __pyx_t_5++; if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_3);
-        #endif
-      }
-    } else {
-      __pyx_t_3 = __pyx_t_6(__pyx_t_7);
-      if (unlikely(!__pyx_t_3)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1334, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_3);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_tree, __pyx_t_3);
-    __pyx_t_3 = 0;
-    __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1334, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_3);
-    if (unlikely(__Pyx_ListComp_Append(__pyx_t_2, (PyObject*)__pyx_t_3))) __PYX_ERR(0, 1334, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_v_all_importances = ((PyObject*)__pyx_t_2);
-  __pyx_t_2 = 0;
-
+1335:         self.feature_importances_ = np.mean(all_importances, axis=0)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_np); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_mean); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = PyTuple_New(1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_INCREF(__pyx_v_all_importances);
-  __Pyx_GIVEREF(__pyx_v_all_importances);
-  PyTuple_SET_ITEM(__pyx_t_2, 0, __pyx_v_all_importances);
-  __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_axis, __pyx_int_0) < 0) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_7, __pyx_t_2, __pyx_t_3); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_1) < 0) __PYX_ERR(0, 1335, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
+1336:         self.feature_importances_ /= self.feature_importances_.sum()  # normalize to add to 1
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_7 = __Pyx_PyObject_GetAttrStr(__pyx_t_2, __pyx_n_s_sum); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_t_2 = NULL;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_7))) {
-    __pyx_t_2 = PyMethod_GET_SELF(__pyx_t_7);
-    if (likely(__pyx_t_2)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_7);
-      __Pyx_INCREF(__pyx_t_2);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_7, function);
-    }
-  }
-  __pyx_t_3 = (__pyx_t_2) ? __Pyx_PyObject_CallOneArg(__pyx_t_7, __pyx_t_2) : __Pyx_PyObject_CallNoArg(__pyx_t_7);
-  __Pyx_XDECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-  __pyx_t_7 = __Pyx_PyNumber_InPlaceDivide(__pyx_t_1, __pyx_t_3); if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_7);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (__Pyx_PyObject_SetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_feature_importances, __pyx_t_7) < 0) __PYX_ERR(0, 1336, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_7); __pyx_t_7 = 0;
-
 1337:
-
+1338:     @staticmethod
-
  __pyx_t_2 = __Pyx_PyObject_CallOneArg(__pyx_builtin_staticmethod, __pyx_t_5); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1338, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_bootstrap, __pyx_t_2) < 0) __PYX_ERR(0, 1339, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1339:     def bootstrap(X, treatment, y, tree):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap = {"bootstrap", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap, METH_VARARGS|METH_KEYWORDS, 0};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_treatment = 0;
-  PyObject *__pyx_v_y = 0;
-  PyObject *__pyx_v_tree = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("bootstrap (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_X,&__pyx_n_s_treatment,&__pyx_n_s_y,&__pyx_n_s_tree,0};
-    PyObject* values[4] = {0,0,0,0};
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  4: values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-        CYTHON_FALLTHROUGH;
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_treatment)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 1); __PYX_ERR(0, 1339, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (likely((values[2] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_y)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 2); __PYX_ERR(0, 1339, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  3:
-        if (likely((values[3] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_tree)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, 3); __PYX_ERR(0, 1339, __pyx_L3_error)
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "bootstrap") < 0)) __PYX_ERR(0, 1339, __pyx_L3_error)
-      }
-    } else if (PyTuple_GET_SIZE(__pyx_args) != 4) {
-      goto __pyx_L5_argtuple_error;
-    } else {
-      values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-      values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-      values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-      values[3] = PyTuple_GET_ITEM(__pyx_args, 3);
-    }
-    __pyx_v_X = values[0];
-    __pyx_v_treatment = values[1];
-    __pyx_v_y = values[2];
-    __pyx_v_tree = values[3];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("bootstrap", 1, 4, 4, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1339, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_4bootstrap(__pyx_self, __pyx_v_X, __pyx_v_treatment, __pyx_v_y, __pyx_v_tree);
-  int __pyx_lineno = 0;
-  const char *__pyx_filename = NULL;
-  int __pyx_clineno = 0;
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_4bootstrap(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_X, PyObject *__pyx_v_treatment, PyObject *__pyx_v_y, PyObject *__pyx_v_tree) {
-  PyObject *__pyx_v_random_state = NULL;
-  PyObject *__pyx_v_bt_index = NULL;
-  PyObject *__pyx_v_x_train_bt = NULL;
-  PyObject *__pyx_v_y_train_bt = NULL;
-  PyObject *__pyx_v_treatment_train_bt = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("bootstrap", 0);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.bootstrap", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_random_state);
-  __Pyx_XDECREF(__pyx_v_bt_index);
-  __Pyx_XDECREF(__pyx_v_x_train_bt);
-  __Pyx_XDECREF(__pyx_v_y_train_bt);
-  __Pyx_XDECREF(__pyx_v_treatment_train_bt);
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__69 = PyTuple_Pack(9, __pyx_n_s_X, __pyx_n_s_treatment, __pyx_n_s_y, __pyx_n_s_tree, __pyx_n_s_random_state, __pyx_n_s_bt_index, __pyx_n_s_x_train_bt, __pyx_n_s_y_train_bt, __pyx_n_s_treatment_train_bt); if (unlikely(!__pyx_tuple__69)) __PYX_ERR(0, 1339, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__69);
-  __Pyx_GIVEREF(__pyx_tuple__69);
-/* … */
-  __pyx_t_5 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_5bootstrap, __Pyx_CYFUNCTION_STATICMETHOD, __pyx_n_s_UpliftRandomForestClassifier_boo, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__70)); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1339, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_codeobj__70 = (PyObject*)__Pyx_PyCode_New(4, 0, 9, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__69, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_bootstrap, 1339, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__70)) __PYX_ERR(0, 1339, __pyx_L1_error)
-
+1340:         random_state = check_random_state(tree.random_state)
-
  __Pyx_GetModuleGlobalName(__pyx_t_2, __pyx_n_s_check_random_state); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1340, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_3 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_random_state); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1340, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_4 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_4 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_4)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_4);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-    }
-  }
-  __pyx_t_1 = (__pyx_t_4) ? __Pyx_PyObject_Call2Args(__pyx_t_2, __pyx_t_4, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_2, __pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1340, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_random_state = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1341:         bt_index = random_state.choice(len(X), len(X))
-
  __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_random_state, __pyx_n_s_choice); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1341, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __pyx_t_5 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1341, __pyx_L1_error)
-  __pyx_t_3 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1341, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __pyx_t_5 = PyObject_Length(__pyx_v_X); if (unlikely(__pyx_t_5 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1341, __pyx_L1_error)
-  __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1341, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_6 = NULL;
-  __pyx_t_7 = 0;
-  if (CYTHON_UNPACK_METHODS && likely(PyMethod_Check(__pyx_t_2))) {
-    __pyx_t_6 = PyMethod_GET_SELF(__pyx_t_2);
-    if (likely(__pyx_t_6)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_2);
-      __Pyx_INCREF(__pyx_t_6);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_2, function);
-      __pyx_t_7 = 1;
-    }
-  }
-  #if CYTHON_FAST_PYCALL
-  if (PyFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, __pyx_t_4};
-    __pyx_t_1 = __Pyx_PyFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_7, 2+__pyx_t_7); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  } else
-  #endif
-  #if CYTHON_FAST_PYCCALL
-  if (__Pyx_PyFastCFunction_Check(__pyx_t_2)) {
-    PyObject *__pyx_temp[3] = {__pyx_t_6, __pyx_t_3, __pyx_t_4};
-    __pyx_t_1 = __Pyx_PyCFunction_FastCall(__pyx_t_2, __pyx_temp+1-__pyx_t_7, 2+__pyx_t_7); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
-    __Pyx_XDECREF(__pyx_t_6); __pyx_t_6 = 0;
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  } else
-  #endif
-  {
-    __pyx_t_8 = PyTuple_New(2+__pyx_t_7); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1341, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    if (__pyx_t_6) {
-      __Pyx_GIVEREF(__pyx_t_6); PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_t_6); __pyx_t_6 = NULL;
-    }
-    __Pyx_GIVEREF(__pyx_t_3);
-    PyTuple_SET_ITEM(__pyx_t_8, 0+__pyx_t_7, __pyx_t_3);
-    __Pyx_GIVEREF(__pyx_t_4);
-    PyTuple_SET_ITEM(__pyx_t_8, 1+__pyx_t_7, __pyx_t_4);
-    __pyx_t_3 = 0;
-    __pyx_t_4 = 0;
-    __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_t_8, NULL); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1341, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __pyx_v_bt_index = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1342:         x_train_bt = X[bt_index]
-
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_X, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1342, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_x_train_bt = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1343:         y_train_bt = y[bt_index]
-
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_y, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1343, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_y_train_bt = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1344:         treatment_train_bt = treatment[bt_index]
-
  __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_treatment, __pyx_v_bt_index); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1344, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_v_treatment_train_bt = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1345:         tree.fit(X=x_train_bt, treatment=treatment_train_bt, y=y_train_bt)
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_fit); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1345, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyDict_NewPresized(3); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1345, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_X, __pyx_v_x_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_treatment, __pyx_v_treatment_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
-  if (PyDict_SetItem(__pyx_t_2, __pyx_n_s_y, __pyx_v_y_train_bt) < 0) __PYX_ERR(0, 1345, __pyx_L1_error)
-  __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1345, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+1346:         return tree
-
  __Pyx_XDECREF(__pyx_r);
-  __Pyx_INCREF(__pyx_v_tree);
-  __pyx_r = __pyx_v_tree;
-  goto __pyx_L0;
-
 1347:
-
+1348:     @ignore_warnings(category=FutureWarning)
-
  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_ignore_warnings); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1348, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_3 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1348, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  if (PyDict_SetItem(__pyx_t_3, __pyx_n_s_category, __pyx_builtin_FutureWarning) < 0) __PYX_ERR(0, 1348, __pyx_L1_error)
-  __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_5, __pyx_empty_tuple, __pyx_t_3); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1348, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_6);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-
+1349:     def predict(self, X, full_output=False):
-
/* Python wrapper */
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds); /*proto*/
-static char __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict[] = "\n        Returns the recommended treatment group and predicted optimal\n        probability conditional on using the recommended treatment group.\n\n        Args\n        ----\n        X : ndarray, shape = [num_samples, num_features]\n            An ndarray of the covariates used to train the uplift model.\n\n        full_output : bool, optional (default=False)\n            Whether the UpliftTree algorithm returns upliftScores, pred_nodes\n            alongside the recommended treatment group and p_hat in the treatment group.\n\n        Returns\n        -------\n        y_pred_list : ndarray, shape = (num_samples, num_treatments])\n            An ndarray containing the predicted treatment effect of each treatment group for each sample\n\n        df_res : DataFrame, shape = [num_samples, (num_treatments * 2 + 3)]\n            If `full_output` is `True`, a DataFrame containing the predicted outcome of each treatment and\n            control group, the treatment effect of each treatment group, the treatment group with the\n            highest treatment effect, and the maximum treatment effect for each sample.\n\n        ";
-static PyMethodDef __pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict = {"predict", (PyCFunction)(void*)(PyCFunctionWithKeywords)__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict, METH_VARARGS|METH_KEYWORDS, __pyx_doc_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict};
-static PyObject *__pyx_pw_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict(PyObject *__pyx_self, PyObject *__pyx_args, PyObject *__pyx_kwds) {
-  PyObject *__pyx_v_self = 0;
-  PyObject *__pyx_v_X = 0;
-  PyObject *__pyx_v_full_output = 0;
-  PyObject *__pyx_r = 0;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict (wrapper)", 0);
-  {
-    static PyObject **__pyx_pyargnames[] = {&__pyx_n_s_self,&__pyx_n_s_X,&__pyx_n_s_full_output,0};
-    PyObject* values[3] = {0,0,0};
-    values[2] = ((PyObject *)((PyObject *)Py_False));
-    if (unlikely(__pyx_kwds)) {
-      Py_ssize_t kw_args;
-      const Py_ssize_t pos_args = PyTuple_GET_SIZE(__pyx_args);
-      switch (pos_args) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        CYTHON_FALLTHROUGH;
-        case  1: values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        CYTHON_FALLTHROUGH;
-        case  0: break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-      kw_args = PyDict_Size(__pyx_kwds);
-      switch (pos_args) {
-        case  0:
-        if (likely((values[0] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_self)) != 0)) kw_args--;
-        else goto __pyx_L5_argtuple_error;
-        CYTHON_FALLTHROUGH;
-        case  1:
-        if (likely((values[1] = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_X)) != 0)) kw_args--;
-        else {
-          __Pyx_RaiseArgtupleInvalid("predict", 0, 2, 3, 1); __PYX_ERR(0, 1349, __pyx_L3_error)
-        }
-        CYTHON_FALLTHROUGH;
-        case  2:
-        if (kw_args > 0) {
-          PyObject* value = __Pyx_PyDict_GetItemStr(__pyx_kwds, __pyx_n_s_full_output);
-          if (value) { values[2] = value; kw_args--; }
-        }
-      }
-      if (unlikely(kw_args > 0)) {
-        if (unlikely(__Pyx_ParseOptionalKeywords(__pyx_kwds, __pyx_pyargnames, 0, values, pos_args, "predict") < 0)) __PYX_ERR(0, 1349, __pyx_L3_error)
-      }
-    } else {
-      switch (PyTuple_GET_SIZE(__pyx_args)) {
-        case  3: values[2] = PyTuple_GET_ITEM(__pyx_args, 2);
-        CYTHON_FALLTHROUGH;
-        case  2: values[1] = PyTuple_GET_ITEM(__pyx_args, 1);
-        values[0] = PyTuple_GET_ITEM(__pyx_args, 0);
-        break;
-        default: goto __pyx_L5_argtuple_error;
-      }
-    }
-    __pyx_v_self = values[0];
-    __pyx_v_X = values[1];
-    __pyx_v_full_output = values[2];
-  }
-  goto __pyx_L4_argument_unpacking_done;
-  __pyx_L5_argtuple_error:;
-  __Pyx_RaiseArgtupleInvalid("predict", 0, 2, 3, PyTuple_GET_SIZE(__pyx_args)); __PYX_ERR(0, 1349, __pyx_L3_error)
-  __pyx_L3_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __Pyx_RefNannyFinishContext();
-  return NULL;
-  __pyx_L4_argument_unpacking_done:;
-  __pyx_r = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict(__pyx_self, __pyx_v_self, __pyx_v_X, __pyx_v_full_output);
-
-  /* function exit code */
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value); /* proto */
-/* … */
-static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_6predict(CYTHON_UNUSED PyObject *__pyx_self, PyObject *__pyx_v_self, PyObject *__pyx_v_X, PyObject *__pyx_v_full_output) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *__pyx_cur_scope;
-  PyObject *__pyx_v_y_pred_ensemble = NULL;
-  PyObject *__pyx_v_df_res = NULL;
-  PyObject *__pyx_v_delta_cols = NULL;
-  PyObject *__pyx_v_i_tr = NULL;
-  PyObject *__pyx_v_treatment_group = NULL;
-  PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1 = 0;
-  PyObject *__pyx_v_tree = NULL;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("predict", 0);
-  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict, __pyx_empty_tuple, NULL);
-  if (unlikely(!__pyx_cur_scope)) {
-    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *)Py_None);
-    __Pyx_INCREF(Py_None);
-    __PYX_ERR(0, 1349, __pyx_L1_error)
-  } else {
-    __Pyx_GOTREF(__pyx_cur_scope);
-  }
-  __pyx_cur_scope->__pyx_v_self = __pyx_v_self;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_self);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_self);
-  __pyx_cur_scope->__pyx_v_X = __pyx_v_X;
-  __Pyx_INCREF(__pyx_cur_scope->__pyx_v_X);
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_v_X);
-/* … */
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_4);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_8);
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_v_y_pred_ensemble);
-  __Pyx_XDECREF(__pyx_v_df_res);
-  __Pyx_XDECREF(__pyx_v_delta_cols);
-  __Pyx_XDECREF(__pyx_v_i_tr);
-  __Pyx_XDECREF(__pyx_v_treatment_group);
-  __Pyx_XDECREF(__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1);
-  __Pyx_XDECREF(__pyx_v_tree);
-  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-  __pyx_tuple__71 = PyTuple_Pack(11, __pyx_n_s_self, __pyx_n_s_X, __pyx_n_s_full_output, __pyx_n_s_y_pred_ensemble, __pyx_n_s_df_res, __pyx_n_s_delta_cols, __pyx_n_s_i_tr, __pyx_n_s_treatment_group, __pyx_n_s_genexpr, __pyx_n_s_genexpr, __pyx_n_s_tree); if (unlikely(!__pyx_tuple__71)) __PYX_ERR(0, 1349, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_tuple__71);
-  __Pyx_GIVEREF(__pyx_tuple__71);
-  __pyx_codeobj__72 = (PyObject*)__Pyx_PyCode_New(3, 0, 11, 0, CO_OPTIMIZED|CO_NEWLOCALS, __pyx_empty_bytes, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_tuple__71, __pyx_empty_tuple, __pyx_empty_tuple, __pyx_kp_s_causalml_inference_tree_uplift_p, __pyx_n_s_predict, 1349, __pyx_empty_bytes); if (unlikely(!__pyx_codeobj__72)) __PYX_ERR(0, 1349, __pyx_L1_error)
-/* … */
-  __pyx_t_3 = __Pyx_CyFunction_New(&__pyx_mdef_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict, 0, __pyx_n_s_UpliftRandomForestClassifier_pre_2, NULL, __pyx_n_s_causalml_inference_tree_uplift, __pyx_d, ((PyObject *)__pyx_codeobj__72)); if (unlikely(!__pyx_t_3)) __PYX_ERR(0, 1349, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_3);
-  __Pyx_CyFunction_SetDefaultsTuple(__pyx_t_3, __pyx_tuple__73);
-  __pyx_t_5 = NULL;
-  if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_6))) {
-    __pyx_t_5 = PyMethod_GET_SELF(__pyx_t_6);
-    if (likely(__pyx_t_5)) {
-      PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_6);
-      __Pyx_INCREF(__pyx_t_5);
-      __Pyx_INCREF(function);
-      __Pyx_DECREF_SET(__pyx_t_6, function);
-    }
-  }
-  __pyx_t_2 = (__pyx_t_5) ? __Pyx_PyObject_Call2Args(__pyx_t_6, __pyx_t_5, __pyx_t_3) : __Pyx_PyObject_CallOneArg(__pyx_t_6, __pyx_t_3);
-  __Pyx_XDECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_3); __pyx_t_3 = 0;
-  if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1348, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-  if (__Pyx_SetNameInClass(__pyx_t_1, __pyx_n_s_predict, __pyx_t_2) < 0) __PYX_ERR(0, 1349, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-/* … */
-struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict {
-  PyObject_HEAD
-  PyObject *__pyx_v_X;
-  PyObject *__pyx_v_self;
-};
-
-
 1350:         '''
-
 1351:         Returns the recommended treatment group and predicted optimal
-
 1352:         probability conditional on using the recommended treatment group.
-
 1353:
-
 1354:         Args
-
 1355:         ----
-
 1356:         X : ndarray, shape = [num_samples, num_features]
-
 1357:             An ndarray of the covariates used to train the uplift model.
-
 1358:
-
 1359:         full_output : bool, optional (default=False)
-
 1360:             Whether the UpliftTree algorithm returns upliftScores, pred_nodes
-
 1361:             alongside the recommended treatment group and p_hat in the treatment group.
-
 1362:
-
 1363:         Returns
-
 1364:         -------
-
 1365:         y_pred_list : ndarray, shape = (num_samples, num_treatments])
-
 1366:             An ndarray containing the predicted treatment effect of each treatment group for each sample
-
 1367:
-
 1368:         df_res : DataFrame, shape = [num_samples, (num_treatments * 2 + 3)]
-
 1369:             If `full_output` is `True`, a DataFrame containing the predicted outcome of each treatment and
-
 1370:             control group, the treatment effect of each treatment group, the treatment group with the
-
 1371:             highest treatment effect, and the maximum treatment effect for each sample.
-
 1372:
-
 1373:         '''
-
 1374:         # Make predictions with all trees and take the average
-
 1375:
-
+1376:         if self.n_jobs != 1:
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1376, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __pyx_t_2 = __Pyx_PyInt_NeObjC(__pyx_t_1, __pyx_int_1, 1, 0); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1376, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_2);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_t_2); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1376, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  if (__pyx_t_3) {
-/* … */
-    goto __pyx_L3;
-  }
-
+1377:             y_pred_ensemble = sum(
-
    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1377, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-
+1378:                 Parallel(n_jobs=self.n_jobs, prefer=self.joblib_prefer)
-
    __Pyx_GetModuleGlobalName(__pyx_t_1, __pyx_n_s_Parallel); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = __Pyx_PyDict_NewPresized(2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_jobs); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_n_jobs, __pyx_t_5) < 0) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_joblib_prefer); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (PyDict_SetItem(__pyx_t_4, __pyx_n_s_prefer, __pyx_t_5) < 0) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1378, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
+1379:                 (delayed(tree.predict)(X=X) for tree in self.uplift_forest)
-
static PyObject *__pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_genexpr(PyObject *__pyx_self) {
-  struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *__pyx_cur_scope;
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("genexpr", 0);
-  __pyx_cur_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *)__pyx_tp_new_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr(__pyx_ptype_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr, __pyx_empty_tuple, NULL);
-  if (unlikely(!__pyx_cur_scope)) {
-    __pyx_cur_scope = ((struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_4_genexpr *)Py_None);
-    __Pyx_INCREF(Py_None);
-    __PYX_ERR(0, 1379, __pyx_L1_error)
-  } else {
-    __Pyx_GOTREF(__pyx_cur_scope);
-  }
-  __pyx_cur_scope->__pyx_outer_scope = (struct __pyx_obj_8causalml_9inference_4tree_6uplift___pyx_scope_struct_3_predict *) __pyx_self;
-  __Pyx_INCREF(((PyObject *)__pyx_cur_scope->__pyx_outer_scope));
-  __Pyx_GIVEREF(__pyx_cur_scope->__pyx_outer_scope);
-  {
-    __pyx_CoroutineObject *gen = __Pyx_Generator_New((__pyx_coroutine_body_t) __pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1, NULL, (PyObject *) __pyx_cur_scope, __pyx_n_s_genexpr, __pyx_n_s_UpliftRandomForestClassifier_pre, __pyx_n_s_causalml_inference_tree_uplift); if (unlikely(!gen)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_cur_scope);
-    __Pyx_RefNannyFinishContext();
-    return (PyObject *) gen;
-  }
-
-  /* function exit code */
-  __pyx_L1_error:;
-  __Pyx_AddTraceback("causalml.inference.tree.uplift.UpliftRandomForestClassifier.predict.genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_r = NULL;
-  __Pyx_DECREF(((PyObject *)__pyx_cur_scope));
-  __Pyx_XGIVEREF(__pyx_r);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-
-static PyObject *__pyx_gb_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_2generator1(__pyx_CoroutineObject *__pyx_generator, CYTHON_UNUSED PyThreadState *__pyx_tstate, PyObject *__pyx_sent_value) /* generator body */
-{
-  PyObject *__pyx_r = NULL;
-  __Pyx_RefNannyDeclarations
-  __Pyx_RefNannySetupContext("genexpr", 0);
-  __pyx_L3_first_run:;
-  if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1379, __pyx_L1_error)
-  if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self)) { __Pyx_RaiseClosureNameError("self"); __PYX_ERR(0, 1379, __pyx_L1_error) }
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_outer_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (likely(PyList_CheckExact(__pyx_t_1)) || PyTuple_CheckExact(__pyx_t_1)) {
-    __pyx_t_2 = __pyx_t_1; __Pyx_INCREF(__pyx_t_2); __pyx_t_3 = 0;
-    __pyx_t_4 = NULL;
-  } else {
-    __pyx_t_3 = -1; __pyx_t_2 = PyObject_GetIter(__pyx_t_1); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = Py_TYPE(__pyx_t_2)->tp_iternext; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1379, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_4)) {
-      if (likely(PyList_CheckExact(__pyx_t_2))) {
-        if (__pyx_t_3 >= PyList_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyList_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1379, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      } else {
-        if (__pyx_t_3 >= PyTuple_GET_SIZE(__pyx_t_2)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_1 = PyTuple_GET_ITEM(__pyx_t_2, __pyx_t_3); __Pyx_INCREF(__pyx_t_1); __pyx_t_3++; if (unlikely(0 < 0)) __PYX_ERR(0, 1379, __pyx_L1_error)
-        #else
-        __pyx_t_1 = PySequence_ITEM(__pyx_t_2, __pyx_t_3); __pyx_t_3++; if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_1);
-        #endif
-      }
-    } else {
-      __pyx_t_1 = __pyx_t_4(__pyx_t_2);
-      if (unlikely(!__pyx_t_1)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1379, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_1);
-    }
-    __Pyx_XGOTREF(__pyx_cur_scope->__pyx_v_tree);
-    __Pyx_XDECREF_SET(__pyx_cur_scope->__pyx_v_tree, __pyx_t_1);
-    __Pyx_GIVEREF(__pyx_t_1);
-    __pyx_t_1 = 0;
-    __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_delayed); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_6 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_tree, __pyx_n_s_predict); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __pyx_t_7 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-      __pyx_t_7 = PyMethod_GET_SELF(__pyx_t_5);
-      if (likely(__pyx_t_7)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-        __Pyx_INCREF(__pyx_t_7);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_5, function);
-      }
-    }
-    __pyx_t_1 = (__pyx_t_7) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_7, __pyx_t_6) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_6);
-    __Pyx_XDECREF(__pyx_t_7); __pyx_t_7 = 0;
-    __Pyx_DECREF(__pyx_t_6); __pyx_t_6 = 0;
-    if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    if (unlikely(!__pyx_cur_scope->__pyx_outer_scope->__pyx_v_X)) { __Pyx_RaiseClosureNameError("X"); __PYX_ERR(0, 1379, __pyx_L1_error) }
-    if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_X, __pyx_cur_scope->__pyx_outer_scope->__pyx_v_X) < 0) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __pyx_t_6 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_empty_tuple, __pyx_t_5); if (unlikely(!__pyx_t_6)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_6);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_r = __pyx_t_6;
-    __pyx_t_6 = 0;
-    __Pyx_XGIVEREF(__pyx_t_2);
-    __pyx_cur_scope->__pyx_t_0 = __pyx_t_2;
-    __pyx_cur_scope->__pyx_t_1 = __pyx_t_3;
-    __pyx_cur_scope->__pyx_t_2 = __pyx_t_4;
-    __Pyx_XGIVEREF(__pyx_r);
-    __Pyx_RefNannyFinishContext();
-    __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
-    /* return from generator, yielding value */
-    __pyx_generator->resume_label = 1;
-    return __pyx_r;
-    __pyx_L6_resume_from_yield:;
-    __pyx_t_2 = __pyx_cur_scope->__pyx_t_0;
-    __pyx_cur_scope->__pyx_t_0 = 0;
-    __Pyx_XGOTREF(__pyx_t_2);
-    __pyx_t_3 = __pyx_cur_scope->__pyx_t_1;
-    __pyx_t_4 = __pyx_cur_scope->__pyx_t_2;
-    if (unlikely(!__pyx_sent_value)) __PYX_ERR(0, 1379, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-  CYTHON_MAYBE_UNUSED_VAR(__pyx_cur_scope);
-
-  /* function exit code */
-  PyErr_SetNone(PyExc_StopIteration);
-  goto __pyx_L0;
-  __pyx_L1_error:;
-  __Pyx_XDECREF(__pyx_t_1);
-  __Pyx_XDECREF(__pyx_t_2);
-  __Pyx_XDECREF(__pyx_t_5);
-  __Pyx_XDECREF(__pyx_t_6);
-  __Pyx_XDECREF(__pyx_t_7);
-  __Pyx_AddTraceback("genexpr", __pyx_clineno, __pyx_lineno, __pyx_filename);
-  __pyx_L0:;
-  __Pyx_XDECREF(__pyx_r); __pyx_r = 0;
-  #if !CYTHON_USE_EXC_INFO_STACK
-  __Pyx_Coroutine_ResetAndClearException(__pyx_generator);
-  #endif
-  __pyx_generator->resume_label = -1;
-  __Pyx_Coroutine_clear((PyObject*)__pyx_generator);
-  __Pyx_RefNannyFinishContext();
-  return __pyx_r;
-}
-/* … */
-    __pyx_t_4 = __pyx_pf_8causalml_9inference_4tree_6uplift_28UpliftRandomForestClassifier_7predict_genexpr(((PyObject*)__pyx_cur_scope)); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_1 = NULL;
-    if (CYTHON_UNPACK_METHODS && unlikely(PyMethod_Check(__pyx_t_5))) {
-      __pyx_t_1 = PyMethod_GET_SELF(__pyx_t_5);
-      if (likely(__pyx_t_1)) {
-        PyObject* function = PyMethod_GET_FUNCTION(__pyx_t_5);
-        __Pyx_INCREF(__pyx_t_1);
-        __Pyx_INCREF(function);
-        __Pyx_DECREF_SET(__pyx_t_5, function);
-      }
-    }
-    __pyx_t_2 = (__pyx_t_1) ? __Pyx_PyObject_Call2Args(__pyx_t_5, __pyx_t_1, __pyx_t_4) : __Pyx_PyObject_CallOneArg(__pyx_t_5, __pyx_t_4);
-    __Pyx_XDECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1379, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-
+1380:             ) / len(self.uplift_forest)
-
    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1380, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_6 = PyObject_Length(__pyx_t_2); if (unlikely(__pyx_t_6 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1380, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_t_2 = PyInt_FromSsize_t(__pyx_t_6); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1380, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    __pyx_t_4 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_2); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1380, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    __pyx_v_y_pred_ensemble = __pyx_t_4;
-    __pyx_t_4 = 0;
-
 1381:         else:
-
+1382:             y_pred_ensemble = sum([tree.predict(X=X) for tree in self.uplift_forest]) / len(self.uplift_forest)
-
  /*else*/ {
-    __pyx_t_4 = PyList_New(0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_2);
-    if (likely(PyList_CheckExact(__pyx_t_2)) || PyTuple_CheckExact(__pyx_t_2)) {
-      __pyx_t_5 = __pyx_t_2; __Pyx_INCREF(__pyx_t_5); __pyx_t_6 = 0;
-      __pyx_t_7 = NULL;
-    } else {
-      __pyx_t_6 = -1; __pyx_t_5 = PyObject_GetIter(__pyx_t_2); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_5);
-      __pyx_t_7 = Py_TYPE(__pyx_t_5)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    }
-    __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-    for (;;) {
-      if (likely(!__pyx_t_7)) {
-        if (likely(PyList_CheckExact(__pyx_t_5))) {
-          if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_5)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_2 = PyList_GET_ITEM(__pyx_t_5, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1382, __pyx_L1_error)
-          #else
-          __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_2);
-          #endif
-        } else {
-          if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_5)) break;
-          #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-          __pyx_t_2 = PyTuple_GET_ITEM(__pyx_t_5, __pyx_t_6); __Pyx_INCREF(__pyx_t_2); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1382, __pyx_L1_error)
-          #else
-          __pyx_t_2 = PySequence_ITEM(__pyx_t_5, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
-          __Pyx_GOTREF(__pyx_t_2);
-          #endif
-        }
-      } else {
-        __pyx_t_2 = __pyx_t_7(__pyx_t_5);
-        if (unlikely(!__pyx_t_2)) {
-          PyObject* exc_type = PyErr_Occurred();
-          if (exc_type) {
-            if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-            else __PYX_ERR(0, 1382, __pyx_L1_error)
-          }
-          break;
-        }
-        __Pyx_GOTREF(__pyx_t_2);
-      }
-      __Pyx_XDECREF_SET(__pyx_v_tree, __pyx_t_2);
-      __pyx_t_2 = 0;
-      __pyx_t_2 = __Pyx_PyObject_GetAttrStr(__pyx_v_tree, __pyx_n_s_predict); if (unlikely(!__pyx_t_2)) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_2);
-      __pyx_t_1 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_1);
-      if (PyDict_SetItem(__pyx_t_1, __pyx_n_s_X, __pyx_cur_scope->__pyx_v_X) < 0) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __pyx_t_8 = __Pyx_PyObject_Call(__pyx_t_2, __pyx_empty_tuple, __pyx_t_1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __Pyx_GOTREF(__pyx_t_8);
-      __Pyx_DECREF(__pyx_t_2); __pyx_t_2 = 0;
-      __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-      if (unlikely(__Pyx_ListComp_Append(__pyx_t_4, (PyObject*)__pyx_t_8))) __PYX_ERR(0, 1382, __pyx_L1_error)
-      __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-    }
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_CallOneArg(__pyx_builtin_sum, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_uplift_forest); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_6 = PyObject_Length(__pyx_t_4); if (unlikely(__pyx_t_6 == ((Py_ssize_t)-1))) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = PyInt_FromSsize_t(__pyx_t_6); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_8 = __Pyx_PyNumber_Divide(__pyx_t_5, __pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1382, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_v_y_pred_ensemble = __pyx_t_8;
-    __pyx_t_8 = 0;
-  }
-  __pyx_L3:;
-
 1383:
-
 1384:         # Summarize results into dataframe
-
+1385:         df_res = pd.DataFrame(y_pred_ensemble, columns=self.classes_)
-
  __Pyx_GetModuleGlobalName(__pyx_t_8, __pyx_n_s_pd); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_DataFrame); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  __pyx_t_8 = PyTuple_New(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __Pyx_INCREF(__pyx_v_y_pred_ensemble);
-  __Pyx_GIVEREF(__pyx_v_y_pred_ensemble);
-  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_v_y_pred_ensemble);
-  __pyx_t_5 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  if (PyDict_SetItem(__pyx_t_5, __pyx_n_s_columns, __pyx_t_1) < 0) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_t_8, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1385, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_v_df_res = __pyx_t_1;
-  __pyx_t_1 = 0;
-
+1386:         df_res['recommended_treatment'] = df_res.apply(np.argmax, axis=1)
-
  __pyx_t_1 = __Pyx_PyObject_GetAttrStr(__pyx_v_df_res, __pyx_n_s_apply); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_GetModuleGlobalName(__pyx_t_5, __pyx_n_s_np); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_5, __pyx_n_s_argmax); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __pyx_t_5 = PyTuple_New(1); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_GIVEREF(__pyx_t_8);
-  PyTuple_SET_ITEM(__pyx_t_5, 0, __pyx_t_8);
-  __pyx_t_8 = 0;
-  __pyx_t_8 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_t_1, __pyx_t_5, __pyx_t_8); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_n_s_recommended_treatment, __pyx_t_4) < 0)) __PYX_ERR(0, 1386, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
 1387:
-
 1388:         # Calculate delta
-
+1389:         delta_cols = [f'delta_{treatment_group}' for treatment_group in self.classes_[1:]]
-
  __pyx_t_4 = PyList_New(0); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1389, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1389, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_5 = __Pyx_PyObject_GetSlice(__pyx_t_8, 1, 0, NULL, NULL, &__pyx_slice__3, 1, 0, 0); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_5);
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_5)) || PyTuple_CheckExact(__pyx_t_5)) {
-    __pyx_t_8 = __pyx_t_5; __Pyx_INCREF(__pyx_t_8); __pyx_t_6 = 0;
-    __pyx_t_7 = NULL;
-  } else {
-    __pyx_t_6 = -1; __pyx_t_8 = PyObject_GetIter(__pyx_t_5); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_7 = Py_TYPE(__pyx_t_8)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1389, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_7)) {
-      if (likely(PyList_CheckExact(__pyx_t_8))) {
-        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_8)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1389, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      } else {
-        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_8)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_5 = PyTuple_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_5); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1389, __pyx_L1_error)
-        #else
-        __pyx_t_5 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_5);
-        #endif
-      }
-    } else {
-      __pyx_t_5 = __pyx_t_7(__pyx_t_8);
-      if (unlikely(!__pyx_t_5)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1389, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_5);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_5);
-    __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_FormatSimple(__pyx_v_treatment_group, __pyx_empty_unicode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_n_u_delta, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1389, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(__Pyx_ListComp_Append(__pyx_t_4, (PyObject*)__pyx_t_1))) __PYX_ERR(0, 1389, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-  }
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  __pyx_v_delta_cols = ((PyObject*)__pyx_t_4);
-  __pyx_t_4 = 0;
-
+1390:         for i_tr in range(1, self.n_class):
-
  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_n_class); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __pyx_t_8 = PyTuple_New(2); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1390, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __Pyx_INCREF(__pyx_int_1);
-  __Pyx_GIVEREF(__pyx_int_1);
-  PyTuple_SET_ITEM(__pyx_t_8, 0, __pyx_int_1);
-  __Pyx_GIVEREF(__pyx_t_4);
-  PyTuple_SET_ITEM(__pyx_t_8, 1, __pyx_t_4);
-  __pyx_t_4 = 0;
-  __pyx_t_4 = __Pyx_PyObject_Call(__pyx_builtin_range, __pyx_t_8, NULL); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  if (likely(PyList_CheckExact(__pyx_t_4)) || PyTuple_CheckExact(__pyx_t_4)) {
-    __pyx_t_8 = __pyx_t_4; __Pyx_INCREF(__pyx_t_8); __pyx_t_6 = 0;
-    __pyx_t_7 = NULL;
-  } else {
-    __pyx_t_6 = -1; __pyx_t_8 = PyObject_GetIter(__pyx_t_4); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1390, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __pyx_t_7 = Py_TYPE(__pyx_t_8)->tp_iternext; if (unlikely(!__pyx_t_7)) __PYX_ERR(0, 1390, __pyx_L1_error)
-  }
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  for (;;) {
-    if (likely(!__pyx_t_7)) {
-      if (likely(PyList_CheckExact(__pyx_t_8))) {
-        if (__pyx_t_6 >= PyList_GET_SIZE(__pyx_t_8)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyList_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1390, __pyx_L1_error)
-        #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        #endif
-      } else {
-        if (__pyx_t_6 >= PyTuple_GET_SIZE(__pyx_t_8)) break;
-        #if CYTHON_ASSUME_SAFE_MACROS && !CYTHON_AVOID_BORROWED_REFS
-        __pyx_t_4 = PyTuple_GET_ITEM(__pyx_t_8, __pyx_t_6); __Pyx_INCREF(__pyx_t_4); __pyx_t_6++; if (unlikely(0 < 0)) __PYX_ERR(0, 1390, __pyx_L1_error)
-        #else
-        __pyx_t_4 = PySequence_ITEM(__pyx_t_8, __pyx_t_6); __pyx_t_6++; if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1390, __pyx_L1_error)
-        __Pyx_GOTREF(__pyx_t_4);
-        #endif
-      }
-    } else {
-      __pyx_t_4 = __pyx_t_7(__pyx_t_8);
-      if (unlikely(!__pyx_t_4)) {
-        PyObject* exc_type = PyErr_Occurred();
-        if (exc_type) {
-          if (likely(__Pyx_PyErr_GivenExceptionMatches(exc_type, PyExc_StopIteration))) PyErr_Clear();
-          else __PYX_ERR(0, 1390, __pyx_L1_error)
-        }
-        break;
-      }
-      __Pyx_GOTREF(__pyx_t_4);
-    }
-    __Pyx_XDECREF_SET(__pyx_v_i_tr, __pyx_t_4);
-    __pyx_t_4 = 0;
-/* … */
-  }
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-
+1391:             treatment_group = self.classes_[i_tr]
-
    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_classes); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1391, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_t_4, __pyx_v_i_tr); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1391, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __Pyx_XDECREF_SET(__pyx_v_treatment_group, __pyx_t_1);
-    __pyx_t_1 = 0;
-
+1392:             df_res[f'delta_{treatment_group}'] = df_res[treatment_group] - df_res[self.control_name]
-
    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_treatment_group); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_cur_scope->__pyx_v_self, __pyx_n_s_control_name); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __pyx_t_5 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_t_4); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-    __pyx_t_4 = PyNumber_Subtract(__pyx_t_1, __pyx_t_5); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_4);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    __pyx_t_5 = __Pyx_PyObject_FormatSimple(__pyx_v_treatment_group, __pyx_empty_unicode); if (unlikely(!__pyx_t_5)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_5);
-    __pyx_t_1 = __Pyx_PyUnicode_Concat(__pyx_n_u_delta, __pyx_t_5); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __Pyx_DECREF(__pyx_t_5); __pyx_t_5 = 0;
-    if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_t_1, __pyx_t_4) < 0)) __PYX_ERR(0, 1392, __pyx_L1_error)
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-
 1393:
-
+1394:         df_res['max_delta'] = df_res[delta_cols].max(axis=1)
-
  __pyx_t_8 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_delta_cols); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  __pyx_t_4 = __Pyx_PyObject_GetAttrStr(__pyx_t_8, __pyx_n_s_max); if (unlikely(!__pyx_t_4)) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_4);
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  __pyx_t_8 = __Pyx_PyDict_NewPresized(1); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_8);
-  if (PyDict_SetItem(__pyx_t_8, __pyx_n_s_axis, __pyx_int_1) < 0) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __pyx_t_1 = __Pyx_PyObject_Call(__pyx_t_4, __pyx_empty_tuple, __pyx_t_8); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __Pyx_GOTREF(__pyx_t_1);
-  __Pyx_DECREF(__pyx_t_4); __pyx_t_4 = 0;
-  __Pyx_DECREF(__pyx_t_8); __pyx_t_8 = 0;
-  if (unlikely(PyObject_SetItem(__pyx_v_df_res, __pyx_n_s_max_delta, __pyx_t_1) < 0)) __PYX_ERR(0, 1394, __pyx_L1_error)
-  __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-
 1395:
-
+1396:         if full_output:
-
  __pyx_t_3 = __Pyx_PyObject_IsTrue(__pyx_v_full_output); if (unlikely(__pyx_t_3 < 0)) __PYX_ERR(0, 1396, __pyx_L1_error)
-  if (__pyx_t_3) {
-/* … */
-  }
-
+1397:             return df_res
-
    __Pyx_XDECREF(__pyx_r);
-    __Pyx_INCREF(__pyx_v_df_res);
-    __pyx_r = __pyx_v_df_res;
-    goto __pyx_L0;
-
 1398:         else:
-
+1399:             return df_res[delta_cols].values
-
  /*else*/ {
-    __Pyx_XDECREF(__pyx_r);
-    __pyx_t_1 = __Pyx_PyObject_GetItem(__pyx_v_df_res, __pyx_v_delta_cols); if (unlikely(!__pyx_t_1)) __PYX_ERR(0, 1399, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_1);
-    __pyx_t_8 = __Pyx_PyObject_GetAttrStr(__pyx_t_1, __pyx_n_s_values); if (unlikely(!__pyx_t_8)) __PYX_ERR(0, 1399, __pyx_L1_error)
-    __Pyx_GOTREF(__pyx_t_8);
-    __Pyx_DECREF(__pyx_t_1); __pyx_t_1 = 0;
-    __pyx_r = __pyx_t_8;
-    __pyx_t_8 = 0;
-    goto __pyx_L0;
-  }
-