|
Abstract:
|
In extending count data models from Poisson regression, Dr. Hilbe and I have written about count data models based on many different distributions including generalized Poisson, heterogeneous dispersion negative binomial, Greene's negative binomial (P), Waring's generalized negative binomial, Famoye's generalized negative binomial, Poisson-inverse Gaussian, double Poisson, and others. We wrote and developed software for zero-inflated versions of these models, and we researched regression models based on truncated versions of distributions. Finally, we investigated regression models appropriate for censored (left censored, right censored, and interval censored) outcomes. In independent work, I investigated models appropriate for heaped outcomes based on mixture distributions and on censored outcomes, and on bivariate models for count data based on bivariate distributions and on copula function approximations. Consistent in all of this work was development of software to allow other researchers to estimate these models. This presentation will demonstrate and catalog count data models while discussing their use in software and discussing their interpretation.
|
