Semiparametric Estimation of Path-Specific Effects in the Presence of Unmeasured Confounding and Exposure-Induced Confounding

• Activity Number: 491
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Caleb Miles*+ and Eric Tchetgen Tchetgen and Ilya Shpitser
• Companies: and Harvard University and University of Southampton
• Keywords: causal inference ; mediation analysis ; path-specific effects ; semiparametric efficient ; multiply robust ; epidemiologic methods

Abstract

Mediation analysis has in recent years received considerable attention in the causal inference literature. Previous literature mainly focused on decomposing a total effect into direct and indirect components mediated by an intermediate variable. Existing results have relied on an assumption of no exposure-induced confounding of the intermediate variable, and generally require no unmeasured confounding for nonparametric identification. In this talk, we relax both sets of assumptions and show that certain path-specific effects remain identified under weaker conditions. We focus on the path specific effect solely mediated by the intermediate in view and not by post-exposure confounders, and we propose a semiparametric, multiply-robust framework of inference, for which we provide stabilizing strategies for near-positivity violations. We illustrate with simulations and an HIV application.