Welcome to causalicp’s documentation!

This is a Python implementation of the Invariant Causal Prediction (ICP) algorithm from the 2016 paper “Causal inference using invariant prediction: identification and confidence intervals” by Jonas Peters, Peter Bühlmann and Nicolai Meinshausen.

At the point of writing, and to the best of my knowledge, the only other publicly available implementation of the algorithm is in the R package written by the original authors.

Installation

You can clone this repo or install the python package via pip:

pip install causalicp

The code has been written with an emphasis on readability and on keeping the dependency footprint to a minimum; to this end, the only dependencies outside the standard library are numpy, scipy and termcolor.

Versioning

The package is still at its infancy and its API is subject to change. However, this will be done with care: non backward-compatible changes to the API are reflected by a change to the minor or major version number,

e.g. code written using causalicp==0.1.2 will run with causalicp==0.1.3, but may not run with causalicp==0.2.0.

License

The implementation is open-source and shared under a BSD 3-Clause License. You can find the source code in the GitHub repository.

Feedback

Feedback is most welcome! You can add an issue in the repository or send an email.