Application of Marginalized Zero-Inflated Models When Mediators Have Excess Zeroes

• Activity Number: 285 - Weighting Methods and Mediation Analysis for Causal Inference
• Type: Contributed
• Date/Time: Tuesday, August 9, 2022 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Andrew M Sims* and Leann Long and Hemant Tiwari
• Companies: University of Alabama at Birmingham and University of Alabama at Birmingham and University of Alabama at Birmingham
• Keywords: Zero-inflation; mediation; marginalized zero-inflated

Abstract

Mediation analysis has become exponentially more popular over the last decade as researchers are interested in establishing mechanistic pathways for intervention. Although methods have increased drastically, there are still limited options for mediation analysis with zero-inflated count variables. No current mediation methods specifically address zero-inflated count mediators. In this paper, we propose an extension of the counterfactual approach to mediation with direct and indirect effects to scenarios where the mediator is a count variable with excess zeroes. We derive effect decomposition utilizing marginalized zero-inflated Poisson model (MZIP) for the mediator model. Our proposed work allows straightforward calculation of direct and indirect effects with respect to the overall population mean for the mediator. We extend this methodology to allow mediator-exposure interactions. We apply this novel methodology to an application and test model performance with simulations comparing proposed MZIP mediator framework to existing Poisson and linear model methods.