Mixtures of Mixed Models: A Bayesian Approach

by Donna L. Mohr.

Abstract: Mixed Models are widely used to model longitudinal data. Potentially, mixtures of mixed models will help represent situations where some subgroups of the subjects follow different patterns of development. Several authors have explored frequentist methods for fitting these mixtures, but these are limited by the requirement that the variance structures be the same across the subgroups. By contrast, a Bayesian method allows us to express the notion that the variance structures should be of about the same magnitude, without requiring them to be exactly equal. By example, I show that this allows us to make some distinctions between groups that are missed by the existing frequentist methods.

Key Words: Mixed Model, Longitudinal Data, Hierarchical Linear Model, Bayesian Analysis, Birth-death process

Author:
Donna L. Mohr, dmohr@unf.edu

Editor: E. G. Tsionas, tsionas@aueb.gr

READING THE ARTICLE: You can read the article in portable document (.pdf) format (153038 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 2089 times since OCTOBER 14, 2008.


Return to the InterStat Home Page.