v0.29.0

PyFixest 0.29

Highlights

• The most important news is that we have overhauled PyFixest's small sample corrections, which now 100% match r-fixest! By default, r-fixest and py-fixest standard errors should now match all of the time. If they don't, please reach out to us, chances are that you've found a bug =) Because it took me a significant amount of time to figure out how fixest handles small sample corrections, we have added a note in which I summarize my understanding here: link.
• We have added support for fully saturated difference-in-differences modeling (the Sun-Abraham approach to event studies), including functions for aggregating treatment effects to the cohort level. You can find examples in the DiD-vignette: link
• This in turn allows us to support @apoorvalal's event-study specification test, which may help you decide if you can get away with a simple two-way fixed effects specification for your event study. If you haven't yet seen Apoorva's paper, you can find it on arxiv.
• After I've failed to run a Gelbach decomposition for a work problem on industry scale data, we've overhauled its internals - the entire decomposition is now computed on sparse matrices, with drastic performance improvements.

All Changes

• Jax benchmarks by @s3alfisc in #784
• Add a jax module by @s3alfisc in #792
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #793
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #800
• update readme by @s3alfisc in #801
• Update feiv_.py by @escherpf in #802
• docs: add escherpf as a contributor for code by @allcontributors in #803
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #804
• Add feature to automatically rename categorical variables in etable by @dsliwka in #798
• update cat_template docs by @s3alfisc in #809
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #810
• Docstring improvements for etable; improve display of qmd docs & small extension for rename_categoricals by @dsliwka in #811
• Add a rename_models arg to plotting methods by @s3alfisc in #813
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #816
• Modified code to make lets-plot an optional dependency. by @elchorro in #818
• docs: add elchorro as a contributor for code by @allcontributors in #819
• Feols: Implement R2 for WLS by @AronNemeth in #815
• docs: add AronNemeth as a contributor for code by @allcontributors in #820
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #824
• Prediction error for non-fe model by @greenguy33 in #573
• Model summary if intercept-only by @leostimpfle in #707
• disable numba jit for code run on jit for full coverage by @s3alfisc in #827
• run extended tests on merge to master by @s3alfisc in #831
• Speed up ci by @s3alfisc in #832
• Multcomp tests by @s3alfisc in #833
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #834
• Fix decomposition bug by @s3alfisc in #836
• Fix formula parsing bug by @DTchebotarev in #839
• docs: add DTchebotarev as a contributor for code by @allcontributors in #841
• Decompose performance by @s3alfisc in #837
• Gelbach optimization by @s3alfisc in #843
• Add R2 within to etable #821 by @AronNemeth in #842
• Dense mm for wild boottest by @s3alfisc in #845
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #847
• Add context capture to prediction with new data by @adamvig96 in #848
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #849
• Add options fixef_k = "full" and fixef_k = "nested" by @s3alfisc in #850
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci in #874
• saturated event study + joint tests for cohort-level heterogeneity by @apoorvalal in #762
• documentation update by @s3alfisc in #877

New Contributors

• @escherpf made their first contribution in #802
• @elchorro made their first contribution in #818
• @AronNemeth made their first contribution in #815
• @DTchebotarev made their first contribution in #839

Full Changelog: v0.27.1...v0.29.0