|
Activity Number:
|
87
- Survival and Longitudinal/Clustered Data Analysis
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Biometrics Section
|
|
Abstract #318953
|
|
|
Title:
|
Spline Linear Mixed Effects Models for Causal Mediation Analysis with Longitudinal Data
|
|
Author(s):
|
Jeffrey M Albert* and Tanujit Dey and Jiayang Sun and Wojbor Woyczynski and Hongxu Zhu and Gregory Powers and Meeyoung Min
|
|
Companies:
|
Case Western Reserve University and Brigham & Women's Hospital and George Mason University and Case Western Reserve University and Case Western Reserve University and Case Western Reserve University and The University of Utah
|
|
Keywords:
|
measurement error;
mediation formula;
potential outcomes;
semiparametric model;
drug abuse
|
|
Abstract:
|
Causal mediation analysis is often of interest when both the mediator and the final outcome are measured repeatedly, but limited research has been done for this situation. Most available methods are based on parametric models and are therefore sensitive to model assumptions. In the present work, we provide a flexible and robust approach to causal mediation analysis for longitudinal data via semiparametric continuous-time models. Specifically, we fit spline linear mixed-effects models to the mediator and to the final outcome in a two-step approach in which a predicted mediator is used as a covariate in the final outcome model. Time-varying natural direct and indirect effects are estimated via an extended mediation formula under a sequential ignorability assumption. Simulation study results are presented to compare properties of alternative mediation effect estimators and delta-method estimates of the standard error. The new approach is applied to harmonized data from two cohort studies to assess attention as a mediator of the effect of prenatal tobacco exposure on externalizing behavior in children.
|
Authors who are presenting talks have a * after their name.
