Regression Models for an Event Count

Chapter 10 introduces regression models for an event count. It begins by defining count data and presenting probability distributions that are commonly associated with count responses. The chapter then considers why OLS is not optimal for these types of responses, and presents instead the Poisson regression model. Truncated, censored, and sample-selected variants of the model are next discussed. Because the Poisson model is rarely adequate for count data, the chapter then discusses several variations on the Poisson model, such as the negative binomial regression model, the zero-inflated Poisson and negative binomial models, and the Poisson hurdle model.