Fig. 2.

Comparison of mean predictive ability (Mean Pred. Ability) for grain yield (a), test weight (b), heading date (c), plant height (d) and powdery mildew resistance (e) using four training population optimization methods: Clustering (Weighted proportion of translocation t2BS:2GS·2GL:2BL in the training population and validation set), PEVmean (training population selected by minimization of the PEV mean in the validation set), PEVmean1 (training population selected by minimization of the PEV of each individual in the validation set) and Random (random training population selection). All methods were evaluated for seven different training population sizes. Error bars represent ± one standard error of the mean