On bivariate Poisson regression models

Abstract

In this paper, we consider estimating the parameters of bivariate and zero-inflated bivariate Poisson regression models using the conditional method. This method is compared with the standard method, which uses the joint probability function. Simulations and real applications show that the two methods have almost identical Akaike Information Criteria and parameter estimates, but the conditional method has a much faster execution time than the joint method. We conducted our computations using the R and SAS package. Our results also indicate that the execution time of SAS is faster than that of R.

Keywords

• Correlated count data
• Conditional modeling
• Bivariate Poisson distribution
• Regression models
• Zero-inflated models