|
Activity Number:
|
24
- Recent Advances in Causal Analyses That Tell the Story of Complex Mediation Mechanisms
|
|
Type:
|
Topic-Contributed
|
|
Date/Time:
|
Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #317592
|
|
|
Title:
|
Post-Treatment Confounding in Causal Mediation Studies: A Cutting-Edge Problem and a Novel Solution via Sensitivity Analysis
|
|
Author(s):
|
Guanglei Hong and Fan Yang* and Xu Qin
|
|
Companies:
|
University of Chicago and University of Colorado Denver and University of Pittsburgh
|
|
Keywords:
|
Causal inference;
Direct effect;
Indirect effect;
Potential outcomes;
Selection Bias;
Weighting
|
|
Abstract:
|
Examples of post-treatment confounding are abundant in causal mediation studies that decompose an average treatment effect into a natural indirect effect and a natural direct effect. Past research has generally considered it infeasible to adjust for such a confounder Z because only Z(1) or Z(0) are observed but not both. Some latest remedies rely on an extra assumption about the ignorability of Z. This study proposes a sensitivity analysis strategy for handling post-treatment confounding and incorporates it into weighting-based causal mediation analysis without extra identification assumptions. The key is to obtain the conditional distribution of Z(0) as a function of the observed values of Z(1) and pretreatment covariates X. The analysis then generates a bound for NIE and that for NDE with adjustment for both Z(1) and Z(0) as well as X. The strategy is suitable for binary and continuous measures of Z. Simulation results reveal major strengths and potential limitations of this new solution. A re-analysis of the National Evaluation of Welfare-to-Work Strategies (NEWWS) Riverside data reveals that the initial analytic results are sensitive to omitted post-treatment confounding.
|
Authors who are presenting talks have a * after their name.
