Bayesian Modelling in Epidemiology
Spatial modelling work has focused on developing methods for solving specific problems of disparate spatial scales or the existence of discontinuities and the methodological issues of ecological bias, measurement error, and interpretation of small-area studies and disease mapping. The use of Bayesian statistical smoothing methods to estimate risk of rare diseases in small areas poses interesting questions of model choice and of potential over smoothing of true excess. Within the methodology programme of SAHSU, we have recently achieved a comprehensive study to give guidelines for interpreting the outcome of disease mapping studies. In an ESRC funded project, we are further developing methodology to combine individual level data measured in surveys with aggregated spatial data with the view to improve ecological inference.
