Robust Estimation of Inverse Probability Weights for Marginal Structural Models

• Activity Number: 152
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Kosuke Imai*+ and Marc Ratkovic
• Companies: Princeton University and Princeton University
• Keywords: causal inference ; longitudinal data ; confouding ; marginal structural models ; propensity score ; covariate balance

Abstract

Marginal structural models (MSMs) are becoming increasingly popular among applied researchers as a tool to make causal inference from longitudinal data. Unlike standard regression models, MSMs can adjust for time-dependent observed confounders while avoiding post-treatment bias. Despite their theoretical appeal, a main practical challenge of MSMs is the difficulty in estimating inverse probability weights. Previous studies have found that MSMs can be highly sensitive to model misspecification of treatment assignment model even when the number of time periods is moderate. The effect of misspecification can propagate across time periods because inverse probability weights used for MSMs are typically based on the product of propensity score estimated separately at each time period. To address this problem, we introduce the Covariate Balancing Propensity Score (CBPS) methodology, which estimates the inverse probability weights such that the resulting covariate balance is optimized. We provide a small scale simulation study, which suggests that the CBPS significantly improves the empirical performance of MSMs by making the treatment assignment model robust to misspecification.