Four structures of data with different degree of genetic connectness and heterogeneity of parameters among herds: 1) heterogeneity of all parameters; 2) genetic means homogeneous and other parameters heterogeneous; 3) heterogeneity of phenotypic parameters and 4) no heterogeneity for herds, were simulated. After prediction of breeding values, the percentage of top animals common to group 1 (animals ranked by the true breeding values) and group 2 (animals ranked by the predicted breeding values) was calculated. For data with heterogeneity of all parameters and 0% of genetic connectness, the percentage of bulls common to the two groups was low and large errors were observed in ranking of top bulls. For 100% of genetic connectness that percentage was larger than 73% and errors in ranking of top bulls were low. For cows and young bulls, even for data with 100% of genetic connectness among herds, the prediction of breeding values was strongly affected by the heterogeneity of all parameters. For the structures with homogeneous genetic means, the percentages of common animals between groups were high and the errors in ranking of top animals were low, even under heterogeneity of the other parameters. These results suggest that heterogeneity of genetic means among herds strongly affects the accuracy of prediction of animals breeding values. However, results from genetic evaluations in the presence of heterogeneity among herds for genetic variance, phenotypic mean and variance were similar to those obtained for data without heterogeneity for these parameters.
connectness; genetic evaluation; selection; simulation
