Files
Abstract
This paper discusses the specification and extimation of random effects count data models. A new multivariate count data model with negative binomial marginals is derived. In contrast to the existing multivariate Poisson model, this model allows for over-dispersion, a phenomenon that is frequently encountered in real data applications. In addition, a multi-factor Poisson model with general individual specific covariance structure is formulated, and an algorithm for estimating the parameters of the model by Simulated Maximum Likelihood is presented.