0.8.0

0.8.0 (2024-06-15)

New:

• N/A

Changed

• migrated from nbdev1 to nbdev2

Fixed:

• fixed dependency issues with scikit-learn and pandas

v0.7.0

0.7.0 (2022-08-02)

New:

• N/A

Changed

• updated dependencies

Fixed:

• N/A

v0.6.0

0.6.0 (2021-10-20)

New:

• Added model_name parameter to CausalBertModel to support other DistilBert models (e.g., multilingual)

Changed

• N/A

Fixed:

• N/A

v0.5.0

0.5.0 (2021-09-03)

New:

• Added support for CausalBert

Changed

• Added p parameter to CausalInferenceModel.fit and CausalInferenceModel.predict for user-supplied propensity scores in X-Learner and R-Learner.
• Removed CV from propensity score computations in X-Learner and R-Learner and increase default max_iter to 10000

Fixed:

• Resolved problem with CausalInferenceModel.tune_and_use_default_learner when outcome is continuous
• Changed to max_iter=10000 for default LogisticRegression base learner

v0.4.0

0.4.0 (2021-07-20)

New:

• N/A

Changed

• Use LinearRegression and LogisticRegression as default base learners for s-learner.
• changed parameter name of metalearner_type to method in CausalInferenceModel.

Fixed:

• Resolved mis-references in _balance method (renamed from _minimize_bias).
• Fixed convergence issues and factored out propensity score computations to CausalInferenceModel.compute_propensity_scores.

v0.3.1

0.3.1 (2021-07-19)

New:

• N/A

Changed

• N/A

Fixed:

• Added sample_size parameter to CausalInferenceModel.evalute_robustness

v0.3.0

0.3.0 (2021-07-15)

New:

• Added CausalInferenceModel.evaluate_robustness method to assess robustness of causal estimates using sensitivity analysis

Changed

• reduced dependencies with local metalearner implementations

Fixed:

• N/A

v0.2.0

0.2.0 (2021-06-21)

New:

• key driver analysis

Changed

• CausalInfererenceModel.fit returns self

Fixed:

• N/A

v0.1.3

0.1.3 (2021-06-17)

New:

• N/A

Changed

• N/A

Fixed:

• version fix

v0.1.2

0.1.2 (2021-06-17)

New:

• N/A

Changed

• Better interpretability and explainability of treatment effects

Fixed:

• Fixes to some bugs in preprocessing