Post date: Apr 30, 2014 3:16:17 PM
Speaker: Dr. Zonghui Hu, Biostaistics Research Branch, NIAID/NIH
Title: Estimation of mean response via effective balancing score
(Authors: Zonghui Hu, Dean Follmann and Naisyin Wang)
Abstract
We introduce the effective balancing score for estimation of the mean response under a missing at random mechanism. Unlike conventional balancing scores, the proposed score is constructed via dimension reduction free of model specification. Three types of such scores are introduced: those that carry the covariate information about the missingness, the response, or both. The effective balancing score leads to consistent estimation with little or no loss in efficiency. Compared to existing estimators, it reduces the burden of model specification and is more robust. It is a near-automatic procedure which is most appealing when high-dimensional covariates are involved. We investigate its asymptotic and numerical properties, and illustrate it with a study of HIV disease.