ABSTRACT
Like the previous chapter, this chapter discusses regression models for response variables (Y) that are modeled using discrete conditional distributions. Unlike the previous chapter, this chapter considers distributions that may be well approximated by using either the Poisson or negative binomial distributions; which are commonly used for count data. Likelihood-based methods are used to distinguish between the two distributions. The chapter concludes with a note on data snooping as regards model selection, and presents a general argument for preregistration as a means to improve generalizability.
