Comparison of Genetic Group and Relationship Methods for Mixed Model Sire Evaluation
B. W. Kennedy, John E. Moxley · Journal of Dairy Science · 1975
Mixed model procedures for sire evaluation were used to evaluate 52 sires by the method of grouping bulls according to year of entry into service and by the method of accounting for relationships among sires.A total of 3288 first lactation 305-day Holstein records of milk yield and composition was used to compare the two methods.Pedigrees of all sires were traced two generations for numerator relationships between the sires.Only 8 sires were not related to any other.Each of the remaining 44 sires was related to at least one other; some were related to as many as 16 others.Four sires were inbred.Rank correlations between sire proofs by the two methods were .88,.89,.91,.91,and .93 for milk, fat, and protein yield, and fat and protein percent.Error variances of prediction of sire proofs by groups averaged 18% and 11% larger than those of relationships for yield and composition traits.Accounting for relationships was particularly effective in reducing error variances of prediction for sires that had few daughters.In one application, the Northeast AI Sire Comparison (2), rather than treating sires as a random sample from a single static population, sires are assigned to fixed groups according to AI stud and time of entry into service (4).However, there has not been agreement on methods of grouping (9), and alternatives to grouping may be desirable.If genetic relationships exist among many of the sires under evaluation, accounting for such relationships theoretically would increase the accuracy of evaluation and could eliminate the need for forming sire groups (4).We investigated genetic relationships among AI sires in a dairy cattle population and compared empirically estimates of sire merit and their error variances of prediction from two applications of mixed-model procedures to AI sire evaluation: one grouped sires according to time of entry into service and an alternative accounted for relationships between sires. METHODS AND PROCEDURESHenderson (4) has discussed application of mixed-model methods to sire evaluation.The following outlines procedures which yield best linear unbiased predictors (BLUP) of sire merit.Let the general mixed-model be