ADMISSIBILITY OF LINEAR ESTIMATES OF REGRESSION COEFFICIENT IN A VARIANCE COMPONENTS MODEL UNDER QUADRATIC LOSS FUNCTION
Xianze Xu · 1993
Consider the variance components model EY=Xβ,cov Y=sum from i=1 to (?) θ_i~2V_i,where X∶n×p and V_i(?)0(i=1,2,…,m) are known,β∈R~p,θ_i~2(?)0 or θ_i~20(i=1,2,…,m) are pa-rameters.Let Sβ be linearly estimable.Under quadratic loss function,sufficient andnecessary conditions of a linear estimate is proved to be (?)-admissible about Sβ,where (?)={LY∶L is a k×n matrix},or {LY+α∶L is a k×n matrix,α∈R~k}.