Empirical Bayes estimation of the parameter vector θ=(β^1,σ^2)' in a multiple linear regression model Y=Xβ+ε is considered, where β is the vector of regreasion coeffcient, ε-N(0,σ^2I) with σ^2 unknown. In this paper, we construct the EB estimators of θ by using the kernel estimation of multivariate density function and its partial derivatives, Uuder some momeut couditions on prior distribution we obtain their asymptotic optimality.
展开▼
