|
Activity Number:
|
444
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
ENAR
|
| Abstract - #307675 |
|
Title:
|
Mediation Analysis Using Random Assignment Interacted with Baseline Covariates as Instrumental Variables
|
|
Author(s):
|
Dylan Small*+
|
|
Companies:
|
University of Pennsylvania
|
|
Address:
|
400 Huntsman Hall, 3730 Walnut St., Philadelphia, PA, 19104,
|
|
Keywords:
|
instrumental variables ; mediation analysis ; causal inference
|
|
Abstract:
|
Randomized trials often involve multi-component treatments. Mediation analysis seeks to "open up" such "black box" treatments and explain their mechanism of action. Traditional approaches to mediation analysis (e.g., Baron and Kenny, 1986) make the strong assumption that mediating variables are randomized within the levels of the baseline randomization variable. Ten Have et al. (2005) introduced a structural mean model approach that makes potentially more tenable assumptions when a covariate is available that predicts the mediating variable. We show how Ten Have et al.'s assumptions can be interpreted in terms of instrumental variables (IVs) and present an IV approach that can be implemented in standard software. We develop a sensitivity analysis for our approach and apply our approach to a trial of an intervention for treating depression in primary care practice.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
