Figure 1

Quality of prediction (average over 20 replicates) for different statistical models with alternative heritabilities (number of QTL=100, random selection). The training set consisted of all the sires (n=100) and other randomly selected individuals (n=900) from generations 1 and 2. Phenotypic information was available for all individuals in generations 1–3. The testing set consisted of 1000 individuals from generation 4. (a) Accuracies; (b) prediction biases: deviation of the regression coefficient (true genetic values on predicted genetic values) from 1; (c) mean squared errors (MSE).