|
Activity Number:
|
156
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #305611 |
|
Title:
|
Analysis of Mixture Random Effects Models for Longitudinal Data
|
|
Author(s):
|
Yimeng Lu*+ and Hongtu Zhu and Thaddeus Tarpey and Eva Petkova
|
|
Companies:
|
Columbia University and Columbia University and New York State Psychiatric Institute and Wright State University and Columbia University
|
|
Address:
|
100 Haven Ave., New York, NY, 10032,
|
|
Keywords:
|
mixture random effects modeling ; Bayesian inference ; MCMC ; antidepressant study
|
|
Abstract:
|
We propose a mixture random effects model to classify subjects based on repeated measurements over time. The hierarchical model accounts for random subject-specific regression coefficients, component-specific distribution of the random effects, and the covariate effect on classification (i.e., on the multinomial probabilities of subjects belonging to given components of the mixture). A Bayesian procedure is developed to estimate component allocation and mean regression coefficients for each component simultaneously. A Markov chain Monte Carlo method is implemented to produce the solution. The procedure is tested through simulations and applied to longitudinal data originated from an antidepressant study with the goal of identifying different types of responders, such as placebo responders or true drug responders.
|
