Probabilistic Evaluation of Null Models for West Nile Virus in the United States

Abstract

AbstractNull models provide a useful baseline for the development of new models. A variety of options for null models exist. These options have become more sophisticated with the advent of probabilistic modeling approaches. Here, we evaluate 10 different null models for West Nile virus, a primarily mosquito-borne disease introduced to the United States in 1999. The Historical Null was significantly better than all models other than the Negative Binomial. We recommend the use of either of these models as a baseline when developing new models to predict spatial and temporal dynamics of West Nile virus at the county-annual scale. We expect these results to be scale-dependent, and a future direction is to examine performance of null models at finer spatial and temporal scales.