v0.12.0b4

This is a beta preparing for our next major release, but does not contain any new user-facing features.

v0.12.0b3

This is a beta preparing for our next major release, but does not contain any new user-facing features.

v0.12.0b2

This is a beta preparing for our next major release, but does not contain any new user-facing features.

v0.12.0b1

This is a beta preparing for our next major release, but does not contain any new user-facing features.

v0.11.1

This is a minor bugfix release. Changes include:

• A fix to forest tuning (#462)
• A new notebook (#466)
• Miscellaneous minor fixes to documentation and code (#468)

v0.11.0

This is a minor release which:

• Extends support for weighting samples to allow both fractional sample weights as well as frequency weights (#439)
• Fixes some problems with the multi-investment case study and improved policy learners (#441)
• Adds a notebook which uses EconML to estimate treatment effects using the dataset from LaLonde (#448)
• Enables pandas dataframes to be used with CausalForestDML, including tuning (#447)
• Fixes a few other miscellaneous issues (#458, #459)

v0.10.0

This release contains a few new features:

• Introduces new classes for policy learning (see DRPolicyTree and DRPolicyForest in our documentation) (#377)
• Exposes the entire set of nuisance models and scores from training when using multiple monte carlo iterations for ortho-learner subclasses (previously only the final ones were kept) (#433)

It also fixes an interoperability issue with DoWhy (#434). Note that this change also removes the deprecated n_splits argument to our estimators, which had already been renamed to cv for the past several releases.

v0.9.2

This is a minor release that adds the following features:

• Enables easy hyperparameter tuning for CausalForestDML (#390)
• Enables CausalForestDML to compute doubly-robust estimates of the ATE and ATT (#391)

v0.9.1

This is primarily a bugfix release; it has the following improvements:

• Reenable using scikit-learn > 0.22.0 but < 0.24.0 (#422)
• Add more robustness to use of feature names with transformers with inconsitent APIs (#422)
• Provide more precise ATE confidence intervals for linear final models (#418)
• A few small miscellaneous improvements (#419)

v0.9.0

This is release contains several major new features as well as a few important breaks in backwards compatibility.

• Introduces Cython implementations of GRF and CausalForestDML, greatly improving the performance of these estimators (#341)

• Enables first stage nuisance estimates to be cached, allowing refitting only the final model for ortho learner subclasses (#360)

• Enables averaging nuisance estimates over several random splits, resulting in lower variance estimates for ortho learner subclasses (#360)

• Adds an RScorer class for performing model selection over different CATE estimators (#361)

• Enables getting SHAP feature importances for CATE estimates (#336, #369)

• More tightly integrates with the dowhy library. For instance, the causal graph used by an estimator can be viewed via est.dowhy.view_model() (#400)

• Improves the display of summaries of inference objects (#407)

• Major Breaking Change: restructured package organization, moving estimators to more consistent locations; for example, the IntentToTreatDRIV estimator is now found at econml.iv.dr.IntentToTreatDRIV. For the moment, we also support using the old package organization (e.g. econml.ortho_iv.IntentToTreatDRIV), but this is deprecated and will be removed in a subsequent release (#370)

• Breaking Change: the n_splits initializer argument for ortho learner subclasses has been renamed to cv to better match sklearn. For the moment, it is still possible to use the name n_splits, but this will be removed in a future release (#362)

• Breaking Change: the base version of the econml package no longer depends on tensorflow or keras (both of which are needed for using DeepIV), or matplotlib (which is needed for rendering tree interpreters). If you need to install these, the first two can be gotten via the econml[tf] extra and matplotlib can be gotten by the econml[plt] extra, or all three libraries can be installed at once via the econml[all] extra (#413).

• Many small fixes and improvements (#337, #358, #373, #363, #365, #328, #398)