Sensitivity Analyses for Parametric Causal Mediation Effect Estimation

• Activity Number: 595
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Jeffrey Albert*+ and Wei Wang
• Companies: Case Western Reserve Univ and Bausch and Lomb, Inc.
• Keywords: causal inference ; interaction ; mediation formula ; potential outcome ; structural equations model

Abstract

Causal mediation analysis uses a potential outcomes framework to estimate the direct effect of an exposure on an outcome and its indirect effect through an intermediate variable (or mediator). Causal interpretations of these effects typically rely on sequential ignorability. Because this assumption is not empirically testable, it is important to conduct sensitivity analyses. Sensitivity analyses so far offered for this situation have focused on the case where the outcome follows a linear model. We propose simple alternative approaches that are suitable as well for generalized linear models. These approaches make use of hybrid structural-observational models that extend the causal model for the outcome and involve easily interpretable sensitivity parameters. The methods are applied to data from a study of the effect of social environmental factors on dental caries in adolescence. We compare two alternative approaches and thus study the sensitivity of the sensitivity analyses.