International genomic evaluations for young bulls

Abstract

Methods were developed to combine MACE solutions of proven bulls with individual GMACE equations for young bulls, to eliminate a double-counting of genomic information previously observed when applying population-based GMACE equations. With the new methods, two options were considered for variance parameters; the same MACE variances for both proven and young bulls, or separate variances for young bulls, estimated from GEBV data. Cross-validation was used to compare how well excluded local GEBV could be predicted from foreign data by each method. All validation results for comparisons among young genomically-tested bulls favoured the new methods and separate variances. Across all traits and scales of evaluation, the correlation with local GEBV increased from .86 for population-based GMACE to .90 for the new methods with separate variances, average absolute differences decreased from 0.13 to 0.06 and root mean squared differences decreased from 0.39 to 0.29. Errors of regression also decreased from 0.17 to 0.09, indicating less bias in comparisons of top versus average young bulls and likely more consistent ranking of top young versus top proven bulls. Approximate reliabilities, and variances of international evaluations were lower with the new methods, as expected with less double-counting of information.