Some Comments on "Bayes" Estimators

Abstract : In recent years, sampling theorists have used certain features of posterior distributions to produce point estimators which are then judged on their properties in repeated samples which have not actually occurred. Some writers even seem to believe that the production of point estimators is the main purpose of Bayes analysis. The authors argue in this note that the concept of the point estimator is of little value; it is an abuse of the Bayesian approach, to use the posterior distribution to produce sampling theory point estimators; and, persistence on this line of enquiry leads to the contradictions and inconsistencies which are to be expected.