Machine learning and the James–Stein estimator

Abstract It is now 62 years since the publication of James and Stein’s seminal article on estimation a multivariate normal mean vector. The paper made spectacular first impression statistical community through its demonstration inadmissability maximum likelihood estimator. continues to be influential, but not for initial reasons. Empirical Bayes shrinkage estimation, major topic, found early justification in James–Stein formula. Less obvious downstream topics include Tweedie’s formula Benjamini Hochberg’s false discovery rate algorithm. This short mainly non-technical review rule effects machine learning era innovation.