Bayesian inference for the negative binomial-Sushila linear model (Q2003101)

This article explores the use of the negative binomial-Sushila (NB-S) distribution in generalized linear models to develop a NB-S linear model for count data. New parametrizations of the distribution are employed. The model parameters are estimated using a Bayesian approach. The comparison of generalized linear models based on Poisson, negative binomial and NB-S distributions to two real data sets with over-dispersion is presented, considering the deviance, the deviance information criterion and the effective number of parameters for model comparison. The NB-S model seems to outperform the other models in this study.