Analysis of medical data by empirical Bayes method

Abstract

The empirical Bayesian analysis of rare rates is considered in the paper. The condition for non-singularity of Bayesian estimates is given and the clustering algorithm is developed, using the property of the Poisson{Gaussian model to treat probabilities of events in populations being the same, if the variance of probabilities is small. The approach developed is applied to the analysis of dissimilarities of homicide and suicide data in Lithuania, 2003-2004.