Search

The animal model for genetic evaluations of dairy cattle by the USDA currently includes a term for interaction effects of sire and herd. The relative magnitude of the variance of that effect was established in the 1960s as 14% of the total variance, but recent research has shown that the proportion is 2% or less. This report compared EBV using either the 14% or the actual estimate from 20 samples of records from herds in California, New York, and Pennsylvania. From 6 to 22% of bulls or cows selected for milk and fat yields based on evaluation with 14% of the total variance would not be selected using the sample estimates, depending on selection intensity, region, and whether only first or up to three lactations were used in the evaluations. Nevertheless, the average EBV of the bulls and cows selected based on 14% of the total variance were only slightly less than for those selected on 2%. This pilot research suggests that further study of the national data be done to establish the appropriate proportion of variance from interaction effects of sire and herd to use with national evaluations. Kinds of evaluations of bulls and ages of cows and bulls should be considered.

An animal model with a REML algorithm was used to estimate variances of additive genetic effects and interaction effects of sire and herd. Milk and fat yields were analyzed for fmt, second, and thrd lactations of Holsteins from California, New York, and Pennsylvania. Twenty samples of data were used in the study: 10 from California, 4 from New York, and 6 from Pennsylvania. Mean number of lactations per sample was 36,820 from 18,189 cows in 156 herds. Mean fractions of phenotypic variance of interaction effects of sire and herd for milk and fat yields were .015 and .019 for first lactation and .019 and .021 for all (up to three) lactations rather than the .14 used for national genetic evaluations in the US. Mean heritability estimates for milk and fat yields were .26 and .24 for first lactation and .21 and .21 for all lactations in California and .34 and .35 for first lactations and .28 and .29 for all lactations in New York. Sums of variances of permanent environmental and interaction effects of sire and herd were similar to those used for national genetic evaluations in the US. Analysis of another 10 samples from California and 10 samples from New York showed only slightly different fractions of phenotypic variance for milk yield for interaction effects of sue by herd, sire by herd by year, and sire by herd by year by season: .023, .027, and .037 for California and .023, .017, and .023 for New York, respectively.

First lactation yields of milk, fat, and protein from Holstein cows in New York and California were used to obtain REML estimates of (co)variances for yield traits using a multitrait animal model. Data from each state were split randomly into 10 samples, averaging 5504 cows per sample from California and 5078 from New York. Mean heritability estimates for milk, fat, and protein yields were .30 ± .02, .31 ± .01, and .29 ± .01 for California data and .33 ± .01, .35 ± .01, and .30 ± .01 for New York data. Averages of genetic correlation estimates for California and New York were .63 ± .01 and .52 ± .02 between milk and fat, .84 ± .01 and -83 ± .01 between milk and protein, and .73 ± .01 and .68 ± .01 between fat and protein. Estimates of environmental correlations were larger than estimates of genetic correlations. Mean estimates of phenotypic correlations for California and New York were .75 ± .01 and .72 ± .01 between milk and fat, .92 ± .01 and .91 ± .01 between milk and protein, and .81 ± .01 and .79 ± .01 between fat and protein yields. On average, these estimates agree with those obtained from animal models with limited rounds of iteration for small data files.