Figure 4

Predicting quality traits from metabolomic composition. (a) The predictive performance of models based on single datasets using OPLS and all datasets together using MB-OPLS. Cross-validation based $r_{\mathsf{\text{CV}}}^2$ statistics equals 1 for perfect predictions. The stars indicate significance level as estimated by the empirical P_CV-value. (b-d) Prediction performance during the median cross-validation run. Grey lines indicate identity.