Regression Models for Count Data from Truncated Distributions

Abstract

We present new commands for analyzing count-data regression models for truncated distributions. The trncregress command allows specification of a regression model for the mean of the truncated distribution through options. In addition to support for truncated Poisson and negative binomial, trncregress fits models based on truncated versions of distributions including generalized Poisson, Poisson-inverse Gaussian, three-parameter negative binomial power, three-parameter Waring negative binomial, and three-parameter Famoye negative binomial.