Table 2.

Mean prediction correlations and bias found with GBLUP over 50 repeats with 7,800 cherry-picked markers, using the three populations separately and combined for modeling.

Prediction set/ training set	MASPOT	Test panel DK	Test panel UK	Combined
DRY MATTER
MASPOT [755]	0.73 [1.01]	0.67 [0.87]	0.59 [1.16]	0.74 [0.93]
Test panel DK [90]	0.72 [1.47]	0.78 [1.34]	0.66 [1.74]	0.79 [0.99]
Test panel UK [289]	0.46 [1.18]	0.44 [1.13]	0.72 [1.29]	0.71 [1.16]
CHIPPING QUALITY
MASPOT [524]	0.51 [1.03]	0.37 [0.75]	0.30 [0.21]	0.49 [0.63]
Test panel DK [42]	0.53 [1.09]	0.26 [1.25]	0.37 [0.25]	0.39 [0.41]
Test panel UK [289]	0.44 [3.04]	0.28 [3.04]	0.79 [1.30]	0.78 [1.39]

The population used for training the model is listed horizontally while the predicted population is listed vertically. Bias is listed in brackets. The number of phenotypes available in each case is indicated with brackets. Bold lettering indicates within-population predictions, where the same population was used for training and test population with 5-fold cross-validation.