Linearly sufficient fun with credibility (Q808604)

let T be a nonrandom matrix. The statistic Tx is linearly sufficient for y if the credibility estimator of y based on x is equal to the credibility estimator of y based on Tx. It is well-known that if a statistic is linearly sufficient in a credibility estimation problem it is usually much simpler to find the coefficients of the credibility estimator. In the present note some techniques for showing that a statistic is linearly sufficient are discussed. In particular, the relation between sufficiency and linear sufficiency is analyzed, and the Bühlmann-Straub model is discussed in the light of linear sufficiency.