Bayes linear estimator in the general linear model

The Bayes linear estimator is obtained by minimizing the Bayes risk matrix under squared loss among all linear estimators.

In this paper, we study the statistical properties and equivalence problems of Bayes linear estimators in the general linear model.

First, we examine linear sufficiency and linear completeness of Bayes linear estimators.

Second, we derive necessary and sufficient conditions under which two Bayes linear estimators coincide.

These conditions clarify when a Bayes linear estimator based on a simpler covariance structure retains Bayes-risk optimality under the original covariance structure.

Moreover, several examples, including Rao's mixed-effects model and the spatial error model, show that our results can simplify the estimation procedure.

Finally, we establish equivalent conditions for the equality of residual sums of squares when Bayes linear estimators are considered.