BEST LINEAR UNBIASED PREDICTOR FOR GENERAL COMBINIG ABILITY AND COMBINED ANALYSIS OF GRIFFING’S DESIGNS TWO AND FOUR

Abstract

The design of a series of experiments which evaluate, in different environments, the same set of diallel crosses is a common test in plant breeding. However, the appropriate model has not been used when data from these experiments are analyzed. This should be a linear mixed model because of the random nature of the general and specific combinig ability effects. In this paper, the combined best linear unbiased predictors (BLUP,s) of general combinig abilities are derived, for the Griffing’s experiments II and IV, established in randomized complete block designs, under the basis of the estimation methodology of random effects in a linear model of mixed effects. A computational algoritm written in SAS/IML commands is also given for the combined analysis of the experiments.