Implementation of Single- and Multiple-Trait Animal Models for Genetic Evaluation of Holstein Type Traits

Animal model genetic evaluations for final score and 14 linear type traits were implemented for US Holsteins in July 1991 and July 1992, respectively. Data included 5.19 million records on 2.83 million cows for final score and 3.52 million records on 2.05 million cows for 14 linear traits. The model included effects of herd-year-month of classification, herd by sire interaction, permanent environment, and animal with unknownparent groups. A multiple-trait procedure for the linear traits was facilitated by a canonical transformation in which the (co)variance matrix for the herd by sire and permanent environment effects were assumed to be proportional to that of the residual effect. Memory requirements were reduced by ordering the data by herds because over 87% of the effects were nested within herds. Investigation of three different computing platforms identified a Unix workstation as the best computing environment. Correlations between the sire and animal model breeding values for bulls with semen available in 1992 were .92 for final score and .91 to .96 for the linear traits. Genetic merit of the mates was accounted for more completely by the animal model, resulting in important differences in breeding value predictions between the two models.