Mixed Effects and Multilevel Models in the Presence of Random Effects Heterogeneity

• Activity Number: 525
• Type: Contributed
• Date/Time: Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): David Afshartous*+ and Geert Verbeke and Richard A. Preston
• Companies: University of Miami and Katholieke Universiteit Leuven and University of Miami
• Keywords: hierarchical linear models ; shrinkage ; random effects ; Gibbs sampling ; Bayes ; variance components

Abstract

We study the impact of a certain type of random effects heterogeneity on the mixed effects model, viz., that in which the variance of the random effects differs between two identifiable subgroups in the population. We extend the basic mixed effects model to account for this type of heterogeneity in the random effects distribution, and then formally examine its effect on the random effects estimates and the subject level Empirical Bayes estimates. We consider three different models in the context of growth curve data from an actual clinical trial. We demonstrate that such heterogeneity may be accommodated in practice via more than one model specification, yielding the same marginal model but with different random effects specifications. Finally, we present the results of a simulation study, focusing on the impact of model misspecification in the presence of random effects heterogeneity.