Online Program Home
  My Program

Abstract Details

Activity Number: 619 - Causal Inference in Biometric Data
Type: Contributed
Date/Time: Thursday, August 3, 2017 : 8:30 AM to 10:20 AM
Sponsor: Biometrics Section
Abstract #324388 View Presentation
Title: Sensitivity Analyzes for Causal Moderation Models for Zero-Inflated Data
Author(s): Robert Gallop* and Paul Crits-Christoph
Companies: and University of Pennsylvania
Keywords: zero-inflated ; causal ; unmeasured confounding ; instrumental variables ; rank preserving models ; distribution free
Abstract:

Within the NIDA Clinical Trials Network (CTN) studies CTN0018 and CTN0019, two primary measures: drug use and sex under the influence, are count responses with a large amount of zeros. Aim of the CTN studies focused on greater effectiveness for the five session intervention (experimental treatment), relative to the one session intervention (standard treatment). Mediation models assess mechanism through which the interventions work. Moderation models assess modifiers of the magnitude of the intervention effect. Standard approaches to mediation and moderation assumes the mediators and moderators are not confounded by unobserved variables. This assumption is not reasonable for any post randomization variable. Hence, the intervention, mediation, and moderation effects, may not be able to be distinguished from the effects of confounders. Our goal is to present an application of causal inference methods for zero-inflated measures in the assessment of mediation and moderation. The combined CTN data provides sufficient power for this investigation. Comparison between the standard methods and causal method for zero-inflated data will be made, with attention on sensitivity analyses.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association