On choice of statistics in credibility estimation
Bjørn Sundt · Scandinavian Actuarial Journal · 1979
The question on what statistics to base our credibility estimators, is discussed in a general model. We introduce concepts of sufficiency, completeness, θ-sufficiency, and θ-completeness that are useful in this connection, and use methods of Rao-Blackwell type. Some of the present results are closely related to results by Taylor (1977).