Bayesian modeling of spatiotemporal patterns of the cumulative incidence of COVID-19 in municipalities of Mexico

Abstract

The work seeks to model the distribution of cumulative incidence rate of COVID-19 in municipalities of Mexico through the adjustment of three generalized linear models (in competition) with spatial and temporal effects and Poisson link function. Data of confirmed cases of COVID-19, reported by Health Secretary of Mexico, from February to July 2020, were used. In order to reduce the computational costs associated with the estimation of multiple parameters with large amounts of data, we chose the Integrated Nested Laplace Approximation implemented in R language (R-INLA). The models were evaluated through the Akaike (AIC) criterion, and the best was the Non-parametric Model of Space-Time Interaction. The results confirm the presence of significant levels of heterogeneity in the spatial-temporal distribution of COVID-19 incidence’s rates among municipalities of Mexico.