Bayesian estimates of disease maps: how important are priors?