Instrumental Variable Approach for Mediation Analysis of Zero-Inflated Count Model

• Activity Number: 526
• Type: Topic Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Zijian Guo*+ and Stuart Gansky and Jing Cheng
• Companies: Statistics Department,Wharton School, University of Pennsylvania and UCSF School of Dentistry and University of California, San Francisco
• Keywords: Instrumental Variable ; Mediation Analysis ; Zero-Inflated Count Model

Abstract

A lot of approaches on mediation analysis focus on continuous or binary mediators and/or outcomes. However, count or zero-inflated count data is common in many studies. In this study, we develop causal methods based on instrumental variable approaches for mediation analysis for those data without the assumption of sequential ignorability when confounders invalidate the assumption. Simulation studies show that our method works well. Our method is applied in a study of caries prevention to study the direct and mediation effects of intervention on count and zero-inflated count outcomes through oral health behaviors.