Joint distribution of the true and estimated errors for two real datasets and different classifier rule models. The training size is 60, and d = 5 features are selected using t-test feature selection method: (a) multiple myeloma, LDA and 5F-CV; (b) multiple myeloma, LDA and LOO; (c) multiple myeloma, LDA and 0.632 bootstrap; (d) HCC, RBF-SVM and 5F-CV; (e) HCC, RBF-SVM and LOO; (f) HCC, RBF-SVM and 0.632 bootstrap. The white line shows the axis, the dotted line shows the regression line and the circle indicates the sample mean of the joint distribution