Prediction methods and heterogeneity effects of residual variances within genetic treatments in clone tests

Abstract

The aim of this study was to compare, through simulation, the BLUP and BLUP-HET procedures for breeding values prediction under heterogeneity of residual variances. Random data were generated with spread sheet, considering a variance of 0,10 and variable residual variance, by number of clones, in order to
provide heterogeneity of variances. Actual and residual breeding values obtained were added to the mean 10, in order to obtain positive phenotypic values. Experimental design was random blocks with 100 genotypes, one plant per plot and 2, 5, 10 and 20 repetitions. Data were evaluated using Selegen software, obtaining estimated breeding values through BLUP and BLUP-HET procedures which were compared to actual
breeding values. In this study, using two and five repetitions showed low accuracy. With heritability close to 10%, it is recommend to use ten or more repetitions in order to ensure greater accuracy in estimates. This represents no restrain in case of variances heterogeneity within genotypes, being both methods suitable. Nevertheless, for most of the cases, BLUP-HET results in accuracies closer to expected values. Furthermore, its estimates for selection gain are closer to real figures.

doi: 10.4336/2011.pfb.31.67.193