A New Multilevel Bayesian Additive Regression Trees for Causal Inference

• In Event: Advances in Multilevel Modeling

Abstract

The ultimate scientific research for causal inferences generally comprises randomized control trials (RCTs). However, random treatment assignment can be unfeasible or unethical. Meanwhile, observational studies can contribute to social science research in many meaningful ways. The current study proposed a new Multilevel Bayesian Additive Regression Trees (BART) algorithm as an alternative strategy to propensity score matching in causal inference. The proposed Multilevel BART algorithm decomposes a multilevel continuous outcome into a fixed and a random component, and then estimate using the BART and a hieratical linear regression model, respectively. In this study, a series of simulation studies were conducted to systematically investigate the performance of Multilevel BART perform under various data scenarios.

Authors

• Siqi Chen, Texas A&M University - College Station

• Wen Luo, Texas A&M University - College Station