Spatially dependent mixture models via the logistic multivariate CAR prior

We consider the problem of spatially dependent areal data, where for each area independent observations are available, and propose to model density through a finite mixture Gaussian distributions. The spatial dependence is introduced via novel joint distribution collection vectors in simplex, that we term logisticMCAR. show salient features logisticMCAR can be described analytically, suitable augmentation scheme based on Pólya-Gamma identity allows derive an efficient Markov Chain Monte Carlo algorithm. When compared competitors, our has proved better estimate densities different (disconnected) locations when they have characteristics. discuss application real dataset Airbnb listings city Amsterdam, also showing how easily incorporate additional covariate information model.