Regression Models for Count Data Based on the Negative Binomial(p) Distribution

Abstract

We present new Stata commands for estimating several regression models suitable for analyzing overdispersed count outcomes. The nbregp command nests the dispersion(constant) and dispersion(mean) versions of Stata's nbreg command in a model for negative binomial(p) regression. The zignbreg command extends Stata's gnbreg command for zero inflation, and the zinbregp command fits a negative binomial(p) regression model with zero inflation. The new commands for zero-inflated models allow specification of links within the glm command's collection for the Bernoulli model of zero inflation. These commands will optionally calculate a Vuong test, which compares the zero-inflated model with the nonzero-inflated model.