Spatio-temporal Bayesian model with two-level spatial structure that includes covariate effect based on mortality data in Nigeria

Abstract

Spatio-temporal models suffer from comparability problems of relative risks (RRs) based on the removal of the covariate effect as a confounding factor on the risk estimate of the study population through distribution of standardized mortality ratios (SMRs). This paper proposed an alternative spatio-temporal Bayesian model with two-level spatial structure that included covariate effect. The objectives were to estimate posterior means for the model hyper-parameters and compare the performance of the two-level spatial structure model implemented with and without the covariate effect using using female breast cancer mortality data extracted from National System of Cancer Registries (NSCR) in Nigeria, from 2009-2016 through Integrated Nested Laplace Approximation (INLA) estimation procedures. The results showed that, the deviance information criterion (DIC) values for the study models were almost identical, and the posterior estimates for the parameters do not change considerably between the two models