Bayesian Modelling for Spatially Misaligned Health Areal Data: A Multiple Membership Approach

Abstract Diabetes prevalence is on the rise in United Kingdom, and for public health strategy, estimation of relative disease risk subsequent mapping important. We consider an application to London data diabetes mortality. In order improve risks, we analyse jointly mortality ensure borrowing strength over two outcomes. The available involve spatial frameworks, areas (Middle Layer Super Output Areas, MSOAs) general practices (GPs) recruiting patients from several areas. This raises a misalignment issue that deal with by employing multiple membership principle. Specifically, translate areal effects explain GP practice according proportions populations resident different A sparse implementation RStan both multivariate conditional autoregressive (MCAR) generalised MCAR (GMCAR) allows comparison these bivariate priors as well exploring implications patterns necessary causal precedence specific conditionality assumption GMCAR, not always present context mapping. Additionally, area-locality considered locate high versus low clusters.