Bayesian Inference of Discrete Weibull Regression Model for Excess Zero Counts

Abstract

This research aimed to study the use of Bayesian estimation for the zero-inflated and hurdle discrete Weibull regression models. Moreover, this study compared the performance of the Bayesian estimation with uniform noninformative priors and informative priors using the random walk Metropolis algorithm and the maximum likelihood estimation. A simulation study was conducted to compare the performance of three different estimation methods by using mean square error with three cases of a simple explanatory variable. A real dataset was analyzed to see how the model works in practice. The results from the simulation study showed that the Bayesian estimation with informative priors is more appropriate for the zeroinflated and hurdle discrete Weibull regression models than other methods. Moreover, the results from a real data application revealed that the Bayes estimators with informative priors for the zero-inflated and hurdle discrete Weibull regression models are the best fitting models.