Skip to Content

Statistics Department Hosted Seminar: Professor Ya'acov Ritov, A priori vs. a posteriori consistency in high dimensional estimation

Tuesday, October 21, 2014
12:00 AM
340 West Hall

We argue that the existence of a consistent estimator with a given rate implies that a Bayes procedure is almost surely consistent with the same rate, where almost surely is measured by the prior. However, we present different examples in which any Bayes estimator which is consistent in the frequentist sense, has tailor made prior which contradicts any a priori belief.