Minimum risk equivariant estimator in linear regression model

Abstract

The minimum risk equivariant estimator of the regression parameter vector is finite-sample optimal, but its calculation is difficult. We study some possible approximations of MRE: A finite-sample approximation uses the Hájek-Hoeffding projection or the Hoeffding-van Zwet decomposition of an initial equivariant estimator.

A large-sample approximation, using the score function of the errors, is based on the asymptotic representation of the initial estimator.

Keywords

• Minimum
• equivariant
• estimator
• linear
• regression
• model