Effect of Heterogeneity of Variance on Dairy Sire Evaluation

The effects of heterogeneity of variance among herds on dairy sire evaluation were examined. Data consisted of 111,196 first lactation Holstein production records of heifers calving in Ontario between 1978 and 1983. To account for selection, older sires without first crop daughters in the data were treated as fixed effects. Herds were classified into one of nine groups based on their estimated sire and residual components. Sires were evaluated by three models. Heterogeneity of variance was ignored in the first model. When accounting for heterogeneity of variance, a genetic correlation of unity between the genetic values of the same sire in the nine variance groups was assumed in one model, and correlations of <1 were assumed for a second model. Data were randomly divided into two equal subsets and sires were evaluated by each method within the subsets. Correlations for the various models were above .92 indicating small differences among the models. Accounting for heterogeneity of variance did not improve the accuracy of sire evaluations.