Risk and Regret of Hierarchical Bayesian Learners

Common statistical practice has shown that the full power of Bayesian methods is not realized until hierarchical priors are used, as these allow for greater "robustness" and the ability to "share stat