Computation Bayesian dynamical mapping of SARS CoV-2 in Sulawesi Tenggara

Risk analysis of SARS CoV-2 is a complex issue since it involves spatial (or location) heterogeneity and actual mobility of people. Empirical Bayes is often used to analyze the relative risk of epidemic cases. This model is expressed as a generalized linear model. In addition, it contains elements of spatial heterogeneity. The SARS CoV-2 is not only affected by spatial heterogeneity but also the actual mobility of people. In other hand, the term uncertainty factor can be expressed as local transmission, global transmission, and their combination. This study develops empirical Bayes, namely compiling several variations of the model based on the uncertainty factors. These models are called local dynamic model, global dynamic model, and intrinsic model (combined local and global dynamic). The daily data of SARS CoV-2 (26 March - 03 August 2020) in 17 districts/cities in Provinsi Sulawesi Tenggara-Indonesia was used to measure the validation of the models. The parameters of these models are estimated using the Bayesian Markov Chain Monte Carlo approach. The estimation result using WinBUGS shows that at iterated 3000 times, burn-in 1000 has converged. This is indicated by the necessary requirements of MCMC, namely irreducibility, aperiodicity and recurency. Intrinsic Bayesian model is the best one because it has the smallest DIC of 21,32. The relative risk mapping using this model shows that Kendari City, Kolaka Utara, Bombana and Buton Tengah are the very highest risk areas of SARS CoV-2 in. Provinsi Sulawesi Tenggara. These four areas have a high potential to spread the SARS CoV-2 to other surrounding area.