Improving Doubly Robust Estimation via Model Comparison

• Activity Number: 152
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Zhiqiang Tan*+
• Companies: Rutgers University
• Keywords: causal inference ; double robustness ; missing data ; propensity score

Abstract

Consider estimating the mean of an outcome in the presence of missing data or estimating population average treatment effects in causal inference. A doubly robust estimator is consistent if an outcome regression model or a propensity score model is correctly specified. In this talk, we present a new doubly robust approach via model comparison. The new approach gives an indication which model is more likely to be correctly specified. Moreover, the resulting estimator is not only as efficient as outcome regression based estimators if the outcome regression model is correctly specified, but also as efficient as existing doubly robust estimators if the propensity score model is correctly specified. Applications to observational studies in medicine and economics will be discussed.