Bayesian Hierarchical Invariant Prediction

TL;DR

BHIP introduces a Bayesian hierarchical approach to invariant causal prediction, enhancing scalability and incorporating prior knowledge, with demonstrated effectiveness on synthetic and real datasets.

Contribution

It reframes ICP using Hierarchical Bayes, improving scalability and enabling prior information integration in causal inference.

Findings

BHIP outperforms ICP in computational scalability.

It effectively tests invariance of causal mechanisms.

Demonstrated success on synthetic and real-world data.

Abstract

We propose Bayesian Hierarchical Invariant Prediction (BHIP) reframing Invariant Causal Prediction (ICP) through the lens of Hierarchical Bayes. We leverage the hierarchical structure to explicitly test invariance of causal mechanisms under heterogeneous data, resulting in improved computational scalability for a larger number of predictors compared to ICP. Moreover, given its Bayesian nature BHIP enables the use of prior information. We evaluate BHIP on both synthetic and real-world datasets, demonstrating its potential as an alternative inference method to ICP and related methods.