|
Activity Number:
|
81
- Contributed Poster Presentations: Section on Statistics in Epidemiology
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2020 : 10:00 AM to 2:00 PM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
|
Abstract #312780
|
|
|
Title:
|
A Semiparametric Modeling Approach to Causal Mediation Analysis for Longitudinal Data
|
|
Author(s):
|
Youjun Li* and Jeffery M Albert
|
|
Companies:
|
Case Western Reserve University and Case Western Reserve University
|
|
Keywords:
|
Causal mediation analysis;
Semi-parametric modeling;
Longitudinal data analysis;
Bayesian inference
|
|
Abstract:
|
Causal mediation analysis provides a tool for exploring possible treatment mechanisms by studying factors lying in the causal pathway between the treatment and the outcome. It has been widely applied in a variety of disciplines due to the scientific interest in not only causal effects but also causal mechanisms. One of the most commonly seen frameworks for causal mediation analysis is the Linear Structural Equation Model (LSEM), under which the Average Causal Mediation Effect (ACME) can easily be obtained by identifying the regression coefficients of the LSEM. In real-life contexts, relationships between the exposure and the presumed mediator as well as the outcome are rarely linear, which implies that LSEM might not be appropriate. Our work aims to handle unbalanced and non-linear longitudinal data using semi-parametric curve fitting under the mediation analysis framework. At the same time, we intend to keep the identification of ACME as simple as the linear case. We will also conduct the same analysis using Bayesian approach to overcome the convergence problem that frequentist approach tends to have.
|
Authors who are presenting talks have a * after their name.
