Stable Weights That Balance Covariates for Causal Inference and Estimation with Incomplete Data

• Activity Number: 386
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Jose Zubizarreta*+
• Companies: Columbia University
• Keywords: causal inference ; observational study ; propensity score ; quadratic programming ; unit nonresponse ; weight adjustment

Abstract

Weighting methods that adjust for observed covariates, such as inverse probability weighting, are widely used for causal inference and estimation with incomplete data. Part of the conceptual appeal of such weighting methods is that one set of weights can be used to estimate a range of treatment effects based on different outcome variables, or that one set of weights can be used to estimate the mean of a variety of variables of interest. However, this appeal is eclipsed by the instability of the estimated weights and by the difficulty in some settings of adequately balancing observed covariates. This paper presents a new weighting method that overcomes these difficulties. Specifically, by solving a quadratic programming problem, this method finds the weights of minimum variance that balance the empirical distribution of the observed covariates up to levels prespecified by the researcher. Duality theory provides useful insight into the behavior of the variance of the optimal weights in relation to the level of covariate balance adjustment. In practice this method is easy to use as currently implemented in the new sbw package in R.