Title: Non-normal estimation of multiple spatial data using multivariate skews normal process
Authors: Kassahun Abere Ayalew - Centers for Disease Control and Preventions (South Africa) [presenting]
Samuel Manda - University of Pretoria (South Africa)
Bo Cai - University of South Carolina (United States)
Abstract: A Multivariate Gaussian Intrinsic Conditional Autoregressive (MICAR-normal) model is used in joint spatial modeling. However, the modelled multivariate data could be highly tailed and skewed. We present a multivariate skew-normal Intrinsic Conditional Autoregressive model (MICAR-skew-normal) to capture the non-normal distribution of the spatial data. We show how to obtain estimates of the model parameters using a fully Bayesian analysis using a stochastic approximation of the EM algorithm (SAEM). Using extensive simulation studies, we demonstrate the capabilities of the proposed model and its usefulness with an analysis of HIV data from South Africa.