Inference on the Treatment Effect in Non-Randomized Pretest-Posttest Studies with Missing Data: An Empirical Likelihood Approach

• Activity Number: 414 - Advances in Estimation Methods
• Type: Contributed
• Date/Time: Tuesday, July 31, 2018 : 2:00 PM to 3:50 PM
• Sponsor: SSC
• Author(s): Shixiao Zhang* and Peisong Han and Changbao Wu
• Companies: University of Waterloo and University of Michigan and University of Waterloo
• Keywords: Auxiliary information; Biased sampling; Empirical likelihood; Multiple robustness

Abstract

Pretest-posttest trials are widely used to study the treatment effect of an intervention. We propose an empirical likelihood (EL) based methodology for both testing and estimation of the treatment effect in non-randomized pretest-posttest studies where the posttest outcomes are subject to missingness. The proposed EL ratio test and the estimation procedure are both multiply robust, in the sense that both allow multiple working models for the propensity score of treatment assignment, the missingness probability and the outcome regression, and the EL ratio test maintains the correct type I error and the estimators are consistent if a certain combination of those multiple working models are correctly specified. The EL ratio test can also be used to construct an EL ratio confidence interval for the treatment effect, which is known in the EL literature to have better coverage than the confidence interval based on the Wald statistic. Simulations are conducted to show the finite-sample performance of the proposed methodology.