A Flexible Spatial Regression Model for Bounded Count Data with Extra Spatial Variation

‎ In the presence of overdispersion‎, ‎the spatial beta-binomial model can be a useful tool for modeling bounded count data‎. ‎However‎, ‎in practical applications‎, ‎additional complexities such as unobserved heterogeneity‎, ‎model misspecification‎, ‎or the presence of skewness and spatial outliers often lead to greater variability than the beta binomial model can capture‎. ‎In this paper‎, ‎we propose a robust extension of this model that provides greater flexibility without increasing interpretational complexity‎. ‎The approach adapts the Beta-2-Binomial (B2B) model to spatial data by introducing a flexible shape parameter capturing excess variation in the data‎. ‎The model accommodates spatial dependence via a latent Gaussian random field‎. ‎Bayesian inference is performed using Markov Chain Monte Carlo algorithms‎. ‎Through an extensive simulation study and application to Loa loa prevalence data from Cameroon and Nigeria‎, ‎we demonstrate the robustness and improved predictive performance of the proposed model compared to existing spatial binomial approaches.‎