Tools for analyzing and quantifying effects of counfounder variables on machine learning model predictions.


Keywords
machine, learning, predictive, modelling, confounding, bias, conditional, independence, confounder, test, conditional-independence, confounders, machine-learning, permutation-test, predictive-modeling
License
CNRI-Python-GPL-Compatible
Install
pip install mlconfound==0.21.3

Documentation

mlconfound

GitHub license GitHub release GitHub CI Documentation Status arXiv GitHub issues GitHub issues-closed Binder

Tools for analyzing and quantifying effects of counfounder variables on machine learning model predictions.

Install

pip install mlconfound

Usage

# y   : prediction target
# yhat: prediction
# c   : confounder

from mlconfound.stats import partial_confound_test

partial_confound_test(y, yhat, c)

Run the quickstart notebook in Binder: Binder

Read the docs for more details.

Documentation Documentation Status

https://mlconfound.readthedocs.io

Citation

T. Spisak, Statistical quantification of confounding bias in predictive modelling, preprint on arXiv:2111.00814, 2021.