Genetic evaluation strategies for multiple traits and countries
Genetic evaluation strategies for multiple traits in multiple countries were studied using simulated data for three lactation traits in each of two importing and two exporting countries. Pedigree-selected young males were progeny-tested within each country. Proven males with higher predicted genetic merits were mated to higher proportions of the female population, and the very best proven males were used internationally. Females were culled at a rate of 33%, based on phenotypic performance in the most recent lactation. Conversion (CNV), multiple-trait across-country evaluation (MACE) and global animal model (GAM) strategies were considered. Base populations were either unselected, or selected with three alternative levels of semen exchange among countries. Selection in base populations included truncation selection of males in all countries, and higher genetic means for the exporting relative to importing countries. With unselected base populations, errors of prediction (ERP) for top (highest predicted genetic merit) bulls were equivalent (P > .10) for MACE and GAM strategies. The GAM had lower ERP than MACE strategies (P < .01) for all bulls, but MACE strategies reduced ERP relative to CNV by 88% of the corresponding reduction for GAM. With selected base populations, MACE strategies had lower ERP than GAM for top and all bulls (P < .01) if semen exchange was low, and equivalent ERP if semen exchange was high (P > .10). The ERP for MACE and CNV strategies were consistently smaller if the strategies allowed multiple, rather than one trait per country. Comparisons of country populations were biased for all strategies with both unselected and selected base populations, due to missing data and implied assumptions about genetic group effects. The biases were largest, and directly responsible for higher ERP, with the GAM strategy if base populations were selected. Modifications to account for selection in MACE eliminated biases (P > .10) if base populations were unselected, and reduced biases by approximately one half if base populations were selected. MACE strategies modified to account for selection always had the lowest ERP among all strategies for all bulls. The modifications to account for selection in MACE are also recommended for national evaluation systems to reduce biases in sub-population comparisons when pedigree data are incomplete.