Characterizations of admissible linear estimators in the linear model

In this paper it is demonstrated that a homogeneous linear estimator for the vector of parameters in a linear regression model is either a general ridge estimator, or a specific linear function of a general ridge estimator which satisfies certain homogeneous linear restrictions. As long as one is not willing to accept these restrictions also for the unknown parameter vector, the latter set of admissible linear estimators is hardly appropriate, thus leaving general ridge estimators as a reasonable class to choose from. (C) 2003 Elsevier Inc. All rights reserved.