Spatio-Temporal Mixture Model for Identifying Risk Levels of COVID-19 Pandemic in Iraq

Abstract

     This paper focuses on choosing a spatial mixture model with implicitly includes the time to represent the relative risks of COVID-19 pandemic using an appropriate model selection criterion. For this purpose, a more recent criterion so-called the widely Akaike information criterion (WAIC) is used which we believe that its use so limitedly in the context of relative risk modelling. In addition, a graphical method is adopted that is based on a spatial-temporal predictive posterior distribution to select the best model yielding the best predictive accuracy. By applying this model selection criterion, we seek to identify the levels of relative risk, which implicitly represents the determination of the number of the model components of all regions over independent time periods. The estimation of parameters and the model selection are both performed in a Bayesian framework. Also, the means of estimated relative risk for the selected mixture model are mapped to give a clearer picture of distributing the disease risks in each district.