ABSTRACT
This chapter discusses regression models for Poisson data. It includes the extension to the problem of overdispersion and its resolution by considering continuous mixtures of Poisson regression models. The chapter introduces nonlinear regression models via one particular model where the data are considered to be normally distributed but the mean response has a nonlinear structure. Count data sometimes exhibit more zeros than expected under a Poisson sampling model. The basic problem of having excess zeros has been termed “zero-inflation.” The generalized linear model (GLM) includes a fairly broad class of regression models such as binary regression, linear regression, and Poisson regression.
