SEMIPARAMETRIC CAUSAL MEDIATION ANALYSIS WITH UNMEASURED MEDIATOR-OUTCOME CONFOUNDING

Abstract

Although exposure can be randomly assigned in studies of mediation effects, direct intervention on the mediator is often infeasible, making unmeasured mediator-outcome confounding possible. We propose a semiparametric identification of natural direct and indirect effects in the presence of unmeasured mediator-outcome confounding by leveraging heteroskedasticity restrictions on the observed data law. For inference, we develop semiparametric estimators that remain consistent under partial misspecifications of the observed data model. We illustrate the proposed estimators using simulations and an application that evaluates the effect of self-efficacy on fatigue among health care workers during the COVID-19 outbreak.