Skip to content

Latest commit

 

History

History
195 lines (159 loc) · 14.7 KB

CHANGELOG.md

File metadata and controls

195 lines (159 loc) · 14.7 KB
Raw

Changelog

2.2.5 (11/7/2024)

  • fix: hot fixes for the extrapolation step + using the presidential margins to infer a ticket splitting estimate in each house / senate race #140

2.2.4 (11/5/2024)

  • fix: truncation can fail catastrophically when % reporting is too low #138

2.2.3 (11/5/2024)

  • chore: adding additional log #135

2.2.2 (11/5/2024)

  • fix: missing est_correction column in VersionedResults DataFrame in the event of bad data #131

2.2.1 (11/1/2024)

  • chore: downgrade botocore and s3transfer as per live team dependency #128

2.2.0 (11/1/2024)

  • chore: condensed logging of non-modeled units #120
  • feat: improvements to margin outlier detection #121
  • feat: extrapolation rule and improvements to called contest handling #122, #123, #124
  • feat: remove min/max during electoral votes estimation #125

2.1.2 (10/24/2024)

  • feat: agg_model_hard_threshold now defaults to True
  • feat: using cross-validation to find the optimal OLS lambda for use in the BootstrapElectionModel is now optional due to the lambda_ model parameter #115

2.1.1 (10/10/2024)

  • fix: allow multiple alpha values passed in to ModelClient.get_national_summary_votes_estimates() and change that method to return a pandas.DataFrame #111

2.1.0 (09/23/2024)

  • fix: model evaluation functions and margin estimand rounding #94
  • chore: updated requirements to their latest versions #95, #103
  • chore: remove duplicate code from the Estimandizer class #96
  • feat: verbose logging of duplicate units #97 and non-modeled units #105
  • fix: rare division by zero when creating normalized_margin #98
  • fix: aggregate model bug #99
  • feat: save aggregate model (national summary) predictions to s3 #100
  • feat: apply race calls to contest-level predictions in addition to national-level ones #101
  • feat: output the mean of the bootstrap as the point prediction #102
  • feat: save unit-level turnout predictions from the bootstrap model to s3 #104
  • feat: distinguish between genuinely unexpected and non-modeled (non-predictive) units #105
  • feat: options to override/control whether or not to allow the model to produce a race call for specified contests #106

2.0.3 (11/06/2023)

  • fix: fix print bug for aggregate model called states error #90
  • chore: add predicted turnout to predictions dataframe #91

2.0.2 (11/02/2023)

  • fix: allow bootstrap model parameters to be of type int as well as float #86
  • fix: pass alpha to national summary client function #87

2.0.1 (10/23/2023)

  • chore: updating all required packages to their latest versions and addressing some warnings that surfaced during testing #81
  • fix: CLI no longer throws an error if aggregates are missing or specified with columns that don't exist in the data #83

2.0.0 (10/13/2023)

  • fix: improved fixed effect features #69
  • fix: with the CLI, model-specific parameters passed in as a dictionary #70
  • fix: additional logic in the CLI to find the .env file #71
  • feat: all models are now subclasses of BaseElectionModel #72
  • feat: ability to create custom estimands #75
  • feat: bootstrap model #76
  • feat: conformal election model uses new faster quantile regression provided by elex-solver #77

1.0.11 (08/14/2023)

  • fix: upgrade to python 3.10 [#65] (#65)
  • feat: generalize parameter checks [#64] (#64)
  • fix: winsorization test error [#62] (#65)
  • feat: add winsorization option [#58] (#58)
  • fix: clean up tox errors [#57] (#57)
  • refactor: remove model settings [#54] (#54)
  • refactor: dynamically create default aggregates [#53] (#53)
  • feat: add better default aggregates [#52] (#52)

1.0.10 (06/07/2023)

  • fix: fixing a fixed effect bug #35
  • chore: updated boto3 version #36
  • feat: move fixed effect creation from CombinedDataHandler to Featurizer #38
  • refactor: removing residual column for nonreporting units #39
  • refactor: remove total voters column #40
  • feat: add regularization to model #42
  • feat: allow selection of fixed effects #43
  • fix: update checking fixed effect input #44
  • fix: rename regularization parameter #46
  • fix: stop unit tests from writing to s3 #48

1.0.9 (03/06/2022)

  • feat: allow model to return conformalization data #32

1.0.8 (01/11/2022)

  • fix: fix overwriting non-reporting lower/upper bounds in multiple prediction interval case #23
  • fix: fix bug when computing fixed effects #27
  • fix: fix overwritting columns when saving conformalization set/bounds to s3 #28

1.0.7 (10/28/2022)

  • fix: fix mape when uncontested historical baseline #18
  • fix: small relative weights for ecos solver #19

1.0.6 (10/05/2022)

  • fix: Gaussian model bug in lower bound of confidence intervals #8
  • fix: save results even with not enough subunits #13
  • feat: write an error message for conformity values #11
  • feat: automate releases #10
  • chore: wrap dataframe in a list to avoid deprecation #12
  • fix: gaussian merge #9
  • fix: release workflow #14
  • feat: add data README #15

1.0.5 (09/20/2022)

  • fix: Update README link to 2020 model #2
  • chore: set up repository #4

1.0.4 (09/12/2022)

  • chore: update codeowners to public news engineering group #101

1.0.3 (09/08/2022)

  • feat: write conformalization data and gaussian bounds to s3 #86

1.0.2 (09/07/2022)

  • fix: use combined data handler to write results #96

1.0.1 (09/06/2022)

  • fix: set write_data as a class function #93

1.0.0 (09/01/2022)

  • feat: implement fixed effects #80
  • feat: add option to save preprocessed data #76
  • feat: create model results class #82
  • feat: implement skewed sampling #83
  • fix: bugs in combined data and historical aggregations #84
  • fix: handle nonreporting unexpected units #85
  • feat: create integration test #87
  • fix: bug in fixed effects where a column is zeroes only #88
  • chore: remove jfrog instructions and update contribution instructions for open source #89

0.0.8 (07/28/2022)

  • chore: rename observed and unobserved to reporting and nonreporting #68
  • chore: add precommit as a workflow #70
  • fix: change default parameters to empty lists #71
  • chore: add unit tests and small fixes #69
  • fix: use new artifactory secret #72
  • chore: update README #73
  • chore: more README updates #74
  • feat: historic run returns state_data #75

0.0.7 (05/13/2022)

  • fix: bug with observed unexpected subunits #65

0.0.6 (05/12/2022)

  • fix: historical election bug #60
  • feat: add configurable unreporting options (drop and zero) #61
  • fix: add 1 to prediction when calculating size of prediction intervals #63

0.0.5 (04/27/2022)

  • chore: update readme and requirements for releases #51
  • feat: add subunits reporting column #52
  • fix: fix test warnings #53
  • fix: rename LiveDataHandler to MockLiveDataHandler #55
  • chore: upgrade .append to .concat #57
  • fix: move random seed to instantiation #56
  • feat: generate estimates for multiple estimands #54

0.0.4 (03/21/2022)

  • chore: updated pandas version and boto3 version #48
  • fix: run historical model with new estimand #46
  • fix: cli allows multiple parameters #45

0.0.3 (02/24/2022)

  • fix: another fix for n_minimum_reporting ([#43])
  • chore: release beta version

0.0.3-beta.0 (02/23/2022)

  • fix: add n_minimum_reporting #40

0.0.2 (02/22/2022)

  • feat: return custom error for not enough subunits reporting #35
  • chore: rename classes and clean up repo #36
  • chore: rename repo to elex-live-model #37

0.0.2-beta.1 (02/14/2022)

0.0.2-beta.0 (02/11/2022)

Initial release (02/03/2022)