Comparison of Linear Estimators Using Pitman's Measure of Closeness

Abstract

A method is given for the comparison of two linear forms of a common random vector under the criterion of Pitman's measure of closeness. Assuming multivariate normality of the random vector, one can determine exact expressions for the closeness probabilities. The applicability of the theory is illustrated on the comparison of ridge regression estimators.

Key Words:

• Least squares
• Ridge regression