Yu, Sheng-Hua - In: Statistics & Probability Letters 77 (2007) 1, pp. 54-62
Consider stochastic effects linear model Y=X[beta]+[epsilon] with E([beta])=A[alpha],Cov([beta])=[sigma]2V1, E([epsilon])=0,Cov([epsilon])=[sigma]2V2, and E([beta][epsilon]')=0, where V1 and V2 are known positive definite matrices, [alpha][set membership, variant]Rk and [sigma]20 are unknown...