|
Activity Number:
|
234
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #309674 |
|
Title:
|
Structural Nested Mean Models for Time-Varying Causal Effect Moderation
|
|
Author(s):
|
Daniel Almirall*+ and Thomas R. Tenhave and Susan Murphy
|
|
Companies:
|
University of Michigan and University of Pennsylvania and University of Michigan
|
|
Address:
|
650 Hidden Valley Club Drive, Ann Arbor, MI, 48104,
|
|
Keywords:
|
Causal effect moderation ; Structural nested mean model ; G-estimator ; 2-stage regression ; bias-variance trade-off ; time-varying covariates
|
|
Abstract:
|
This talk considers the problem of assessing effect moderation in longitudinal settings in which treatment is time-varying and so are the covariates said to moderate its effect. The main challenges of assessing causal effect moderation in the time-varying setting are discussed. Intermediate Causal Effects that describe time-varying causal effects of treatment conditional on past covariate history are introduced and considered as part of Robins' Structural Nested Mean Model. Two estimators of the causal effects are presented: The first is a 2-Stage Regression Estimator, which can be used with standard regression software. The second is Robins' G-Estimator. The methodology is illustrated using longitudinal data from the PROSPECT study. Our goal is to estimate the effects of time-varying adherence to the intervention conditional on time-varying covariates that may modify these effects.
|
