ABSTRACT
We now consider regression models for count data, where the counts have no clear upper bounds. Specifically we consider models in which the response has a Poisson distribution or a mixture of Poisson distributions. Our focus is on models with linear structure. Section 1 considers Poisson regression. Section 2 introduces the problem of over-dispersion and uses mixture models including zero-inflated Poisson models to deal with over-dispersion. Section 3 briefly discusses longitudinal Poisson modeling.
