A note on spatial partition models , with application toBayesian mapping of disease

We consider the problem of mapping the risk from a disease using a series of regional counts of observed and expected cases. To analyse this problem from a Bayesian viewpoint we propose a methodology, which extends Markov random elds priors by including a partially exchangeable set of parameters. Such an extension allows detection of clusters between remote regions, reeecting some underlying causes of disease. The methodology can be implemented to obtain, by means of Markov chain Monte Carlo sampling, smoothed Bayesian estimators of the relative risk of developing a disease in a certain region.