R. Bras. Zootec.01/Jul/2002;31(4):1680-8.
Effects of Heterogeneity of Residual Variance among Contemporary Groups on Genetic Evaluation of Beef Cattle
DOI: 10.1590/S1516-35982002000700010
The objective of this study was to investigate, via simulated data, the effect of heterogeneity of residual variance among contemporary groups (CG) on genetic evaluation of beef cattle, and to compare a weighted genetic evaluation procedure (R¹Ise2) with one that assumes homogeneity of variances (R = Ise2). Different data sets of postweaning weight gain, adjusted to 345 days, were simulated with and without heterogeneity of residual variance, using a phenotypic variance of 300 kg2 and a true heritability of 0.4. A real data set was used to provide the CG and parents related to each animal observation. Results showed that, when high levels of heterogeneity of residual variance were considered, animals were selected from CG with higher variability, especially with intense selection. With respect to prediction consistency, non parent animals and cows had their predicted breeding values more affected by heterogeneity of residual variance than sires. The weighed factor used reduced, but did not eliminate, the effect of heterogeneity of residual variance. The results of weighted genetic evaluations were similar or superior to those from evaluations that assumed homogeneity of variances. Even when the variances were homogeneous, the weighed genetic evaluations yielded results that were not inferior than those from the usual evaluations, which assumed homogeneity of variances.
Keywords: accuracy; consistency; simulation; weighed estimate