Pedigree information to predict genetic merit for type of Holstein bulls

Three methods of computing predicted differences for type were compared for accuracy of pedigree evaluation of young Holstein bulls. Computing methods were herdmate comparison, herdmate comparison adjusted for genetic merit of herdmates, and best linear unbiased prediction. Evaluations by progeny test were available for 1298, 1293, and 1088 sons and sires for herdmate comparison, herdmate comparison adjusted for cow index of herdmates, and best linear unbiased prediction with fewer sons having more complete pedigrees. Ten regression models were used to predict eventual progeny test evaluations of sons from combinations of evaluations for sires, dams, and maternal grandsires. Negative intercepts and smaller than expected regression coefficients suggested bias in evaluations by all methods but much less bias with best linear unbiased prediction than with other methods. Multiple correlations generally were largest for equations using evaluations by best linear unbiased prediction.