Linear minimax estimation in the three parameters case

Rok, strany: 1999, 311 - 318

We consider the linear model $Ey = Xβ, cov y = I$ under the restriction $|β|≤1$. The Minimax-$B'B$-estimator is the estimator $hatβ = Cy$ minimizing the maximal $B'B$-mean square error. The solution heavily depends on the dimension $j$ of maximal eigenspace of eigenspace of $B(CX-I) (CX-I)'B'$. If $k=3$, only the case $j=2$ poses essential problems. It is shown that if $X'X$ and $B'B$ possess a joint eigenvector, a simple solution of the Linear Minimax Problem may be possible.

ISO 690:

Drygas, H. 1999. Linear minimax estimation in the three parameters case. In Tatra Mountains Mathematical Publications, vol. 17, no.3, pp. 311-318. 1210-3195.

APA:

Drygas, H. (1999). Linear minimax estimation in the three parameters case. Tatra Mountains Mathematical Publications, 17(3), 311-318. 1210-3195.