ABSTRACT
When the response is an unbounded count (0,1,2,3,...), we can use a count regression model to explain this in terms of the given predictors. Sometimes other models may be appropriate. When the count is bounded, a binomial-type response regression as discussed in the previous chapters is sensible. In some cases, the counts might be sufficiently large that a normal approximation is justified so that a normal linear model may be used. We shall consider two distributions for counts: the Poisson and, less commonly, the negative binomial.
