Improving the Finite-Sample Performance of Doubly Robust Estimators Through Focused Nuisance Parameter Estimation

• Activity Number: 152
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Social Statistics Section
• Author(s): Karel Vermeulen*+ and Stijn Vansteelandt
• Companies: Ghent University and Ghent University
• Keywords: Causal inference ; Double robustness ; Finite-sample bias ; Nuisance parameter ; Semi-parametric estimation

Abstract

Doubly robust estimators for the marginal treatment effect are asymptotically unbiased when one of two nuisance working models are correctly specified, regardless of which. While the choice of estimators of the nuisance parameters indexing these models does not affect the asymptotic distribution of such doubly robust estimators when all working models are correctly specified, it can have a dramatic impact under model misspecification. In this talk, we will thus focus on estimation strategies for these nuisance parameters. In particular, we will propose a simple and generic estimation principle for the nuisance parameters indexing each of the working models, which is designed to improve the finite-sample performance of the doubly robust estimator of interest, relative to the common use of maximum likelihood estimators. Our approach improves the stability of the weights in those doubly robust estimators which invoke inverse probability weighting. It moreover results in doubly robust estimators with easy-to-calculate asymptotic variance. Simulation studies confirm the desirable finite-sample performance of the proposed estimators relative to other proposals.