Poisson and Negative Binomial Regression

Communication scholars have used Poisson and negative binomial regression models to study elements of political and health communication, interpersonal and relational communication, as well as mass communication and society. This chapter addresses Poisson and negative binomial regression, two techniques used in analyzing count data. As a generalized linear model (GLM), Poisson regression contains a log link function, a Poisson random component, and one or more independent variables as systematic components. Negative binomial regression is a type of GLM, and like Poisson regression, it is characterized by a log link function as well as a systematic component consisting of categorical and/or continuous explanatory variables. The chapter uses data from the 2008 American National Election Study to demonstrate both Poisson and negative binomial regression techniques in SPSS. In SPSS, the GLMs procedure fits both Poisson and negative binomial regression models; to choose between the two, a researcher must determine whether a dependent measure is overdispersed.