Generalized least squares estimators in the analysis of covariance structures

Abstract

Let S represent the usual unbiased estimator of a covariance matrix. ?o, whose elements are functions of a parameter vector ?o:?o = ?(?o). A generalized least squares (G.L.S.) estimate, ?. of ?o may be obtained by minimizing tr[ {S - ?(?) }V]2 where V is some positive definite matrix.