Variable Selection for Causal Inference via Outcome-Adaptive Random Forest

TL;DR

This paper introduces the outcome-adaptive random forest (OARF), a novel method for variable selection in causal inference that improves bias reduction and variance control in high-dimensional, non-linear settings.

Contribution

The paper presents OARF, a new approach that selectively includes variables related to the outcome for propensity score estimation, enhancing causal effect estimation accuracy.

Findings

OARF produces unbiased estimates with smaller variance.

OARF outperforms existing methods in variable selection accuracy.

Empirical examples show tighter confidence intervals and plausible variable selection.

Abstract

Estimating a causal effect from observational data can be biased if we do not control for self-selection. This selection is based on confounding variables that affect the treatment assignment and the outcome. Propensity score methods aim to correct for confounding. However, not all covariates are confounders. We propose the outcome-adaptive random forest (OARF) that only includes desirable variables for estimating the propensity score to decrease bias and variance. Our approach works in high-dimensional datasets and if the outcome and propensity score model are non-linear and potentially complicated. The OARF excludes covariates that are not associated with the outcome, even in the presence of a large number of spurious variables. Simulation results suggest that the OARF produces unbiased estimates, has a smaller variance and is superior in variable selection compared to other approaches. The results from two empirical examples, the effect of right heart catheterization on mortality and the effect of maternal smoking during pregnancy on birth weight, show comparable treatment effects to previous findings but tighter confidence intervals and more plausible selected variables.