Estimating mortality rates that are unstable due to local variation, is a typical problem of small area statistics. A simple Bayesian method is to assume that the number of deaths in an area is distributed as a binomial variate and construct a prior distribution on its parameters, but it is of limited value. Similarity of mortality rates is greater if the two small areas are geographically closer, so we need to incorporate the relative strength of the similarity. The author proposes a Bayesian method that answers this problem by introducing a general binary regression model assuming smoothness for the response function.
