Fig. 4.

Cross-validation predictive ability (mean of n = 25 twofold cross-validations) in individual populations (Pop I–V) for seven HA traits and DTH. For each population, predictive ability is based on genomic prediction models developed from a multi-population (MP) training set, using the KGD statistical method. Prediction accuracies by KGD in the full MP training set (Fig. 3), are represented here (black symbol) for comparison. Error bars are SE