Table 2 Confusion matrices on a test set of 100 examples, \(\log (B_{10})\), prediction accuracy (PA), Matthews Correlation Coefficient (MCC), F1-Score (F1), \(\kappa \)-Statistic and Youden’s statistic (J) for different settings. (a) always predicting the dominant class, (b) errors only in the dominant class, (c) perfect prediction, (d) random prediction

From: Statistical independence for the evaluation of classifier-based diagnosis