Mostly Harmless Fixed Effects Regression in Python with PyFixest [PyCon DE & PyData Berlin 2024]

PyFixest is a Python port of the R Fixest library for regression analysis, providing high-performance fixed effects regression functionality

The library offers significant performance improvements - up to 5-10x faster than alternatives, being on par with Julia implementation and close to R’s Fixest performance

Key features include:

• OLS regression with high-dimensional fixed effects
• Poisson regression
• Instrumental variable models
• Advanced inference techniques (cluster robust, wild cluster bootstrap)
• Multiple estimation capabilities
• Post-processing options for visualization and result comparison

Uses Wilkinson formulas for intuitive model specification syntax similar to R

Core algorithms are implemented in C++ and parallelized for performance optimization

Automatically handles practical considerations like:

• Dropping multicollinear variables
• Computing robust standard errors
• Providing sensible defaults

Built with compatibility in mind - results match exactly with R’s Fixest implementation

Designed to be user-friendly for R Fixest users - maintains similar API and behavior

Can handle large datasets efficiently (tested with 10M+ observations)

Current limitations include:

• No GLS/WLS support for Poisson regression
• Performance may degrade with very high-dimensional categorical features
• Some IV diagnostics still missing

Active development continues with focus on expanding features and improving code quality