Methods of variance component estimation

Abstract

Estimation of variance components is a method often used in population genetics and applied in animal breeding. Even experienced population geneticists nowadays feel lost if confronted with the huge set of different methods of variance component estimation. Especially because there exists no uniformly best method, a decision which method should be used is often difficult to take. This paper gives a complete overview of methods existing in the simplest case of a one-way lay-out and demonstrates some of them by a numerical example for which the true situation is known. Of course, the one-way lay-out is of limited practical interest but can best be used to explain animal scientists the basic principles of the methods. These basic principles are principally also valid for higher classifications. Advantages and disadvantages of the methods are discussed. The symbols used are the standard biometric symbols as given in Rasch et al. (1994). We can say that all the methods offered by SPSS can be recommended.