A geometrical explanation of Stein shrinkage

DOI10.1214/11-STS382zbMATH Open1330.62282arXiv1203.4737OpenAlexW2001096624MaRDI QIDQ2634652FDOQ2634652

Authors: Lawrence Brown, Linda H. Zhao 

Publication date: 18 February 2016

Published in: Statistical Science (Search for Journal in Brave)

Abstract: Shrinkage estimation has become a basic tool in the analysis of high-dimensional data. Historically and conceptually a key development toward this was the discovery of the inadmissibility of the usual estimator of a multivariate normal mean. This article develops a geometrical explanation for this inadmissibility. By exploiting the spherical symmetry of the problem it is possible to effectively conceptualize the multidimensional setting in a two-dimensional framework that can be easily plotted and geometrically analyzed. We begin with the heuristic explanation for inadmissibility that was given by Stein [In Proceedings of the Third Berkeley Symposium on Mathematical Statistics and Probability, 1954--1955, Vol. I (1956) 197--206, Univ. California Press]. Some geometric figures are included to make this reasoning more tangible. It is also explained why Stein's argument falls short of yielding a proof of inadmissibility, even when the dimension, $p$, is much larger than $p=3$. We then extend the geometric idea to yield increasingly persuasive arguments for inadmissibility when $pgeq3$, albeit at the cost of increased geometric and computational detail.

Full work available at URL: https://arxiv.org/abs/1203.4737

Recommendations

• Stein's method for geometric approximation
• Shrinkage to smooth non-convex cone :Principal component analysis as stein estimation
• Reviving some geometric aspects of shrinkage estimation in linear models
• scientific article; zbMATH DE number 863804
• A unified and generalized set of shrinkage bounds on minimax Stein estimates
• Geometric ergodicity for Bayesian shrinkage models
• Stein's method for compound geometric approximation

zbMATH Keywords

shrinkageminimaxempirical Bayeshigh-dimensional geometryStein estimation

Mathematics Subject Classification ID

Ridge regression; shrinkage estimators (Lasso) (62J07) Admissibility in statistical decision theory (62C15)

Cites Work

• Estimation of the mean of a multivariate normal distribution
• Title not available (Why is that?)
• Title not available (Why is that?)
• Estimation with quadratic loss.
• Title not available (Why is that?)
• On Minimax Statistical Decision Procedures and their Admissibility
• On the Admissibility of Invariant Estimators of One or More Location Parameters
• Stein estimation: The spherically symmetric case
• Title not available (Why is that?)
• Admissible Estimators, Recurrent Diffusions, and Insoluble Boundary Value Problems
• Limiting the Risk of Bayes and Empirical Bayes Estimators--Part I: The Bayes Case
• Admissibility in statistical problems involving a location or scale parameter
• The Admissibility of Certain Invariant Statistical Tests Involving a Translation Parameter

Cited In (4)

• Geometry and applied statistics
• Shrinkage to smooth non-convex cone :Principal component analysis as stein estimation
• Prediction and calibration for multiple correlated variables
• Larry Brown's work on admissibility

This page was built for publication: A geometrical explanation of Stein shrinkage

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2634652)