BEST LINEAR UNBIASED FAMILIAR PREDICTOR FOR PARTIAL DIALLEL EXPERIMENTS WITHOUT MATERNAL EFFECTS

Abstract

When an experiment of diallel crosses involves a subset of all possible simple crosses, which can be generated from a basic set of p
parents, a partial diallel experiment is obtained. The most useful of these experiments are those in which every parent is involved in a number s<p of crosses. If, in addition, the parents are grouped into two or more subsets or families, where the members of a particular subset are distinguished for some common characteristics, then the problem arises when estimating the effects of family or group. This problem is solved in this paper, by means of an empirical best linear unbiased predictor of the family effects, in abscence of maternal effects. As a complement of this research, a computational algorithm in SAS/IML commands is also obtained, in order to analyse the experiments, giving at the same time the corresponding estimators of the family effects.