Adjustment for Heterogeneity of Variances by Herd Production Level in Dairy Cow and Sire Evaluation

Abstract

Additive genetic and environmental variances for all lactation milk yield and natural log of yield were estimated at three herd production levels by REML in a sire and nested-cow model. Data consisted of 121,136 first and later lactations for 91,206 Holstein cows from 526 sires. Each herd-year-season group was assigned to one of three production levels by mean yield of all cows freshening in the same herd-year. Variance components for untransformed yields increased with production level. Heritability was .18, .22, and .24 for untransformed yield in low, medium, and high producing herds. Log transformation did not stabilize variances. Environmental variances for transformed yields decreased as production level increased. Heritability estimates were not significantly changed by log transformation of yields. Cows and sires were evaluated by three models. Models 1 and 2 used variances from the medium level and either untransformed or logtransformed yields. Model 3 used untransformed yields and the specific variances at the three production levels. Rank correlations of sire and cow evaluations from the three models were close to unity. Ranks of top sires were similar, but differences among ranks of top cows were large across models. If heterogeneity of variances is ignored, untransformed yields should be used.