pybdiff 0.1.0

Python replication of Stata's bdiff command — tests whether regression coefficients differ across two subgroups

pybdiff

Python replication of Stata's bdiff command. Tests whether regression coefficients differ across two subgroups using permutation tests, bootstrap tests, or parametric Wald tests — with optional parallel execution.

Attribution

This package is a Python reimplementation of the Stata command bdiff originally written by:

Lian, Yujun (连玉君). bdiff: Coefficient Difference Test across Two Groups. Version 1.04, 24 Nov 2020. Sun Yat-sen University. Email: [e-mail] · Blog: https://www.lianxh.cn

All methodological credit belongs to the original author. The bootstrap and permutation procedures follow:

Efron, B., Tibshirani, R., 1993. An Introduction to the Bootstrap. Chapman & Hall.

Installation

pip install pybdiff

Quick Start

from pybdiff import bdiff

result = bdiff(
    df      = df,
    group   = "treated",          # column with values 0 and 1
    formula = "y ~ x1 + x2 | firm + year",
    vcov    = {"CRV1": "firm"},
    method  = "permutation",      # "permutation", "bootstrap", or "wald"
    reps    = 500,
    seed    = 42,
    n_jobs  = -1,                 # use all CPU cores
)
print(result)

Methods

Method	Type	Description
permutation	Non-parametric	Randomly shuffles group labels (default)
bootstrap	Non-parametric	Resamples each group with replacement
wald	Parametric	Chi-squared test using block-diagonal VCV

Parameters

Parameter	Type	Default	Description
df	pd.DataFrame	—	Dataset with a 0/1 group column
group	str	—	Name of the grouping variable
formula	str	—	PyFixest formula, e.g. "y ~ x1 + x2 | fe"
vcov	str | dict	"iid"	Variance-covariance type
method	str	"permutation"	Test method
reps	int	500	Resampling iterations (ignored for Wald)
seed	int | None	42	Master random seed
n_jobs	int	1	Parallel workers (-1 = all cores)
verbose	bool	True	Print progress bar and results table

Returns

pd.DataFrame indexed by variable name with columns:

• b_group0, b_group1 — estimated coefficients per group
• diff — coefficient difference (b0 − b1)
• stat — test statistic (frequency count or chi-squared)
• p_value — two-sided p-value
• se_diff — standard error of the difference (Wald only)
• valid_reps — successful resampling iterations (resampling only)

Dependencies

• pyfixest
• pandas, numpy, scipy, joblib, tqdm

License

MIT