fix weights issue on gam and bam, also added test case for weights

pymgcv/gam.py

@@ -933,7 +933,7 @@ def fit(
             data=data_to_rdf(data, include=self.referenced_variables),
             family=self.family.rfamily,
             method=method,
-            weights=ro.NULL if weights is None else np.asarray(weights),
+            weights=ro.NULL if weights is None else to_rpy(np.asarray(weights)),
             optimizer=to_rpy(np.array(optimizer)),
             scale=0 if scale is None else (-1 if scale == "unknown" else scale),
             select=select,
@@ -1177,7 +1177,7 @@ def fit(
                 data=data_to_rdf(data, include=self.referenced_variables),
                 family=self.family.rfamily,
                 method=method,
-                weights=ro.NULL if weights is None else np.asarray(weights),
+                weights=ro.NULL if weights is None else ro.FloatVector(np.asarray(weights)), #issue: can't use to_rpy(np.asarray(weights))
                 scale=0 if scale is None else (-1 if scale == "unknown" else scale),
                 select=select,
                 gamma=gamma,

tests/gam_test_cases.py

@@ -102,7 +102,7 @@ def mgcv_gam(
         with ro.local_context() as env:
             env["data"] = data_to_rdf(data, include=self.gam_model.referenced_variables)
             for k, v in self.add_to_r_env.items():
-                env[k] = v
+                env[k] = ro.FloatVector(v) if isinstance(v, np.ndarray) else v  #issue: can't use to_rpy(v)
             result = ro.r(self.mgcv_call)
             assert isinstance(result, ro.ListVector)
             return result
@@ -554,6 +554,27 @@ def family_test_cases() -> dict[str, GAMTestCase]:
         ),
     }
 
+def weighted_gam(gam_type: type[AbstractGAM]) -> GAMTestCase:
+    """test case for GAM with observation weights."""
+    data = get_test_data()
+    n = len(data)
+
+    rng = np.random.default_rng(42)
+    weights = rng.uniform(0.5, 2.0, size=n)
+
+    method = "REML"
+
+    return GAMTestCase(
+        mgcv_args=f"y ~ s(x), weights=w_arr, method='{method}'",
+        gam_model=gam_type({"y": S("x")}),
+        data=data,
+        add_to_r_env={"w_arr": weights},
+        fit_kwargs={
+            "weights": weights,
+            "method": method
+        }
+    )
+
 
 def get_test_cases() -> dict[str, GAMTestCase]:
     supported_types_and_cases = [
@@ -580,6 +601,7 @@ def get_test_cases() -> dict[str, GAMTestCase]:
                 linear_functional_tensor_2d_gam,
                 markov_random_field_neighbours,
                 markov_random_field_polys,
+                weighted_gam
             ],
         ),
     ]