This is a major update to EconML, with new features, support for newer versions of Python and for updated versions of our dependencies, and many bugfixes. The biggest new feature is the addition of a method for sensitivity analysis for several of our estimators - see validation.rst.

What's Changed

New features

• Add sensitivity analysis methods by @fverac in #967, #975
• Alternative scoring metrics by @carl-offerfit in #988
• Add support for numpy 2, Cython 3, Python 3.13 by @kbattocchi in #984
  • Note that since Python 3.8 is no longer supported, we no longer support DeepIV, which relied on an outdated version of TensorFlow. We hope to add a pytorch replacement in the future, see #87.

Bug fixes, minor enhancements, and build improvements

• Fix documentation publishing workflow by @kbattocchi in #894
• Fix ATE bounds frequency by @kgao in #899
• Disable Python 3.13 when building wheels for CI by @kbattocchi in #909
• Switch from pycodestyle to ruff by @kbattocchi in #901
• fix bug in check_high_dimensional by @itamarfaran in #906
• add compiled .c files to .gitignore by @vojavocni in #911
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #910
• Update README information by @kbattocchi in #916
• Fix coverage upload by @kbattocchi in #923
• "Fixes linear DRIV shape inconsistency" by @ssemov in #920
• Add LKG generation to CI by @kbattocchi in #803
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #917
• Enable Discrete Outcomes w/ RScorer by @jakepenzak in #927
• Correct "Help Wanted" url in readme by @KomiB-Atsou in #934
• fix dowhy_ not existance issue by @louis-she in #929
• Enable support for dowhy 0.12 by @kbattocchi in #935
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #928
• Miscellaneous fixes by @kbattocchi in #939
• Fix numpy iterable import in _causal_analysis.py for numpy>2.0 by @imatiach-msft in #945
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #937
• Update supported shap versions by @kbattocchi in #948
• Fix typos in the Customer Segmentation case study by @KuanHaoHuang in #957
• update datasets host url for notebooks by @kgao in #959
• minor readme fix by @fverac in #974
• Bugfix rscorer discrete outcome treatment by @maartenvanhooftds in #977
• allow missing values in X when inferencing for specific ests by @fverac in #982
• ENH: _check_sample_weight dtype made kwarg in sklearn 1.7.x by @atharva-novi in #980

New Contributors

• @itamarfaran made their first contribution in #906
• @vojavocni made their first contribution in #911
• @pre-commit-ci made their first contribution in #910
• @ssemov made their first contribution in #920
• @jakepenzak made their first contribution in #927
• @KomiB-Atsou made their first contribution in #934
• @louis-she made their first contribution in #929
• @KuanHaoHuang made their first contribution in #957
• @maartenvanhooftds made their first contribution in #977
• @atharva-novi made their first contribution in #980
• @carl-offerfit made their first contribution in #988

Full Changelog: v0.15.1...v0.16.0