Causal Inference Using Bayesian Nonparametric Modeling

• Activity Number: 100
• Type: Invited
• Date/Time: Monday, August 1, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Jennifer Hill*+
• Companies: New York University
• Keywords: causal inference ; Bayesian ; propensity score ; missing data ; nonparametric ; common support

Abstract

Researchers have long struggled to identify causal effects in non-experimental settings. Many recently proposed strategies assume ignorability of the treatment assignment mechanism and require fitting two models-one for the assignment mechanism and one for the response surface. An alternate strategy is proposed that instead focuses on very flexibly modeling just the response surface using a Bayesian nonparametric modeling procedure, Bayesian Additive Regression Trees (BART). BART has advantages with regard to ease of use, ability to handle a large number of predictors, coherent uncertainty intervals, and natural accommodation of continuous treatment variables and missing data for the outcome variable. BART also naturally identifies heterogeneous treatment effects. BART has been shown to produce more accurate estimates of average treatment effects compared competitors in the propensity score world in the nonlinear simulation situations examined. Further, it is highly competitive in linear settings with the "correct" model, linear regression. The approach can also be used to identify areas that lack common support. Extensions that allow for covariate missing data will be discussed.