
Title
Model selection and model checking for generalized linear mixed models
Speaker
Hal Stern, Department of Statistics, Iowa State University
Abstract
Random effects are often used in the context of generalized linear models to account for correlation among units in the study. Assumptions about the population distribution of random effects are difficult to verify. This talk focuses on two different questions: 1) determining whether the data support the need for random effects; 2) assessing the fit of the model. For the former we review approaches to computing the Bayes factor for comparing two models. For the latter, we apply the posterior predictive approach to model assessment.