Bayesian Approaches to Model Overdispersion in Spatio-Temporal Binomial Data

In this work, we introduce a direct spatio-temporal extension of the spatial conditional overdispersion models for binomially distributed response variables. This proposal incorporates a spatial term similar to the spatial lag of the response variable for each time unit within the linear predictor. These models effectively capture both spatial and temporal correlations inherent in the dataset under study. Furthermore, we introduce temporally varying spatial lag coefficient models, enabling for the possibility of introducing temporal changes in the spatial term. In order to be able to assess the usefulness of our proposals, we apply them to the analysis of low birth weight in Georgia, providing a comparative analysis of the performance of our models to that of the commonly used Knorr-Held’s models.