Modelando dados de contagem com inflação de zeros, sobredispersão e dependência espacial

Abstract

This work proposes a new model for count data with excess zeros, overdispersion and spatial dependence. To accommodate these characteristics simultaneously, we used an zeroinflated quasi-likelihood (QIZ), where the spatial dependence is incorporated in the estimation process through generalized estimating equations (GEE). The estimation algorithm used in this process was the ES (Expectation-Solution); confidence intervals for the parameters were obtained via Bootstrap Inference. Simulation studies have been performed in various scenarios. Finally, the method is illustrated using data of leprosy cases in the State of Amazonas.