A unified Bayesian approach for modeling zero-inflated count and continuous outcomes

This article reexamines zero-inflated count and semi-continuous models for analyzing data exhibiting an excess of zeros. Most of these models seem to share a common structure belonging to the exponential dispersion family (EDF) of distributions and the two-part hurdle model. When examining cross-sectional outcomes with a distribution belonging to the EDF, several hurdle models have been explored. This includes recently utilized models as well as some new models that are described in detail here. Then, a unified Markov Chain Monte Carlo (MCMC) method is presented for analyzing data with outcomes belonging to the EDF. Furthermore, a user-friendly R package called UHM (unified hurdle models) has been developed and made available on the Comprehensive R Archive Network (CRAN). This package enables users to easily obtain Bayesian estimates of parameters of interest for hurdle models. Finally, the methods developed in this study are applied to analyze two real datasets featuring count and continuous outcomes with a high prevalence of zero values. Additionally, simulation studies are performed to demonstrate and assess the performance of the proposed models.