``A Sensitivity Analysis for Missing Outcomes under the Matched-Pairs Design.''



The matched-pairs design enables researchers to efficiently infer causal effects from randomized experiments. However, like any experiments, the validity of empirical findings can be compromised in the presence of missing outcomes. In this paper, we develop a sensitivity analysis for missing outcomes by exploiting the key feature of the matched-pairs design. The idea is that if two nearly identical observations are paired prior to the randomization of the treatment, the missingness of one unit's outcome is informative about the potential missingness of the other unit's outcome. We consider the average treatment effect among always-observed pairs (AOP) whose units exhibit no missing outcome regardless of their treatment status. The naive estimator based on available pairs is unbiased for the AOP if two units of the same pair are identical in terms of their missingness patterns. The proposed sensitivity analysis characterizes how the bounds of the AOP widen as the degree of the within-pair similarity decreases. We further extend the methodology to the matched-pairs design in observational studies. Our simulation studies show that informative bounds can be obtained under some scenarios when the proportion of missing data is not too large. The proposed methodology is also applied to the randomized evaluation of the Mexican universal health insurance program. (Last Revised, May 2017)

© Kosuke Imai
 Last modified: Sat May 6 12:44:51 EDT 2017