ABSTRACT
In the previous chapter, we introduced generalized linear models (GLMs) that are useful when the outcome variable of interest is categorical in nature. We described a number of models in this broad family, including logistic regression for binary, ordinal, and multinomial data distributions along with Poisson regression models for count or frequency data. In the examples given, the data were collected at a single level. However, just as is true for normally distributed outcome variables, it is common for categorical variables to be gathered in a multilevel framework. The focus of this chapter is on models designed specifically for scenarios in which the outcome of interest is either categorical or counted and the data have been collected in a multilevel framework. Chapter organization will mirror that of Chapter 7. We will start with a description of fitting logistic regression for dichotomous data, followed by models for ordinal and nominal dependent variables. The chapter will conclude with models for frequency count data that fit the Poisson distribution and overdispersed counts.
