Matching on Double Balancing Scores for Estimating Average Treatment Effect in a Small Sample

• Activity Number: 141
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Xiaoshan Wang*+ and Jacqueline Starr
• Companies: Forsyth Institute and Forsyth Institute
• Keywords: Matching estimator ; Propensity score ; Doubly robust ; Prognostic score

Abstract

We are searching for an improvement to the propensity score matching estimator by using double balancing scores. Theoretical aspects of matching on double scores is examined. Different combinations of scores and of matching methods are explored. We show that, by combining balancing scores, the balancing effect can be significantly improved by certain matching method when compared to matching on the propensity score only. The improvement leads to a smaller bias of the estimated average treatment effect. When all models for balancing scores are correctly specified, a moderate sample size can ensure an excellent performance of a double-score matching estimator which is comparable to the doubly robust estimator. A correct choice of matching method with an appropriate combination of balancing scores can also attain certain robust property similar to the doubly robust estimator.