Posterior Simulation and Model Choice in Longitudinal Generalized Linear Models

This paper is concerned with the problems of posterior simulation and model choice for Poisson panel data models with multiple random effects. Efficient algorithms based on Markov Chain Monte Carlo methods for sampling the posterior distribution are developed. A new parameterization of the random effects and fixed effects is proposed and compared with a parameterization in common use. Computation of marginal likelihoods and Bayes factors from the simulation output is also considered. The methods are illustrated with several real data applications involving large samples and multiple random effects