Pretest Estimation of Eigenvalues of a Wishart Matrix

Abstract

In this paper, we discuss various large sample estimation techniques for eigen values of Wishart matrix. Pretest estimators are proposed and these are compared with the sample eigen values using asymptotic quadratic distributional risk. The relative dominance picture of the proposed estimators is presented. A criteria for the selection of the size of pretest is discussed. It is shown that the significance level for the proposed pretest estimator often coincides with the commonly used level of significance.