The Stein Effect for Frechet Means

TL;DR

This paper extends the James-Stein shrinkage estimator to the estimation of Frechet means in non-positively curved metric spaces, showing it can outperform unbiased estimators in high dimensions.

Contribution

It introduces a generalized James-Stein estimator for Frechet means in metric spaces with non-positive curvature, demonstrating asymptotic dominance over unbiased estimators.

Findings

The generalized James-Stein estimator asymptotically dominates unbiased estimators in non-positively curved spaces.

Simulation results confirm the estimator's efficacy on metric trees and positive-definite matrices.

Results extend James-Stein estimator applicability beyond Euclidean spaces to complex metric spaces.

Abstract

The Frechet mean is a useful description of location for a probability distribution on a metric space that is not necessarily a vector space. This article considers simultaneous estimation of multiple Frechet means from a decision-theoretic perspective, and in particular, the extent to which the unbiased estimator of a Frechet mean can be dominated by a generalization of the James-Stein shrinkage estimator. It is shown that if the metric space satisfies a non-positive curvature condition, then this generalized James-Stein estimator asymptotically dominates the unbiased estimator as the dimension of the space grows. These results hold for a large class of distributions on a variety of spaces - including Hilbert spaces - and therefore partially extend known results on the applicability of the James-Stein estimator to non-normal distributions on Euclidean spaces. Simulation studies on metric trees and symmetric-positive-definite matrices are presented, numerically demonstrating the efficacy of this generalized James-Stein estimator.