Nonparametric Bayesian Estimation of Heterogeneous Causal Effects Using Real-World Data

• Activity Number: 315 - Innovative Bayesian Approaches in Clinical Trials and Practical Considerations
• Type: Invited
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xinyi Xu* and Bo Lu and Steve MacEachern and Ling Wang
• Companies: The Ohio State University and The Ohio State University and The Ohio State University and Michigan State University
• Keywords: causal inference ; heterogenous treatment effects; propensity score; Bayesian nonparametrics; Gaussian process

Abstract

Inferring a causal relationship is an important task in social science and health research. In a large population, different subgroups of individuals might respond differently to certain treatments. Identifying and estimating the heterogeneous effects can help researchers improve treatments or better allocate resources to meet the needs. In studies with real world data, propensity score is often used as a dimension reduction tool to aid the inference under the ignorable treatment assignment assumption. We propose a nonparametric Bayesian approach that utilizes propensity scores and observable factors to capture heterogeneous treatment effects. We show that our model produces estimators that take the same form as traditional matching estimators under certain prior specifications, and outperform the matching estimators with improved efficiency and better identification of heterogeneous effects. Furthermore, we apply our method to investigate the impact of college attendance on women fertility, which is known to suffer from the potential heterogeneous effects.