Mixture Model Framework Facilitates Understanding of Zero-Inflated and Hurdle Models for Count Data

In this note, we comment on the zero-inflated and hurdle models for count data presented by Rose et al., 2006, J. Biopharma. Stat. 16:463–481. By viewing these models as finite mixture models, one gains a better understanding of the components of the models, including assumptions about the latent variable(s) in the finite mixture models. Deciding whether a zero-inflated or hurdle model is appropriate for a given data set requires close collaboration with subject matter experts. For instance, in modeling vaccine adverse event count data, the pharmacokinetic rationale for the occurrence of an adverse event and the likelihood of detecting or reporting the adverse event are important considerations for mixture model development.