Stochastic Tree Ensembles for Estimating Heterogeneous Effects

TL;DR

This paper introduces a more efficient algorithm for Bayesian Causal Forests, improving posterior exploration and interval estimation in causal inference, with validation through simulations and empirical analysis.

Contribution

It develops a novel, more efficient algorithm for fitting BCF models, enhancing computational performance and posterior coverage.

Findings

Improved efficiency over previous Gibbs sampler

Better posterior exploration and interval coverage

Validated through simulations and empirical data

Abstract

Determining subgroups that respond especially well (or poorly) to specific interventions (medical or policy) requires new supervised learning methods tailored specifically for causal inference. Bayesian Causal Forest (BCF) is a recent method that has been documented to perform well on data generating processes with strong confounding of the sort that is plausible in many applications. This paper develops a novel algorithm for fitting the BCF model, which is more efficient than the previously available Gibbs sampler. The new algorithm can be used to initialize independent chains of the existing Gibbs sampler leading to better posterior exploration and coverage of the associated interval estimates in simulation studies. The new algorithm is compared to related approaches via simulation studies as well as an empirical analysis.