Fig. 5.
Calibration curves and rugs of predictive probabilities for the colon and prostate cancer data by Dettling (2004). The colon cancer data set consists of 2000 genes with 62 (22 normal and 40 tumor) samples and the prostate cancer data set consist of 6033 genes with 102 (50 normal and 52 tumor) samples. Predictive probability of each sample was computed based on the rest samples as the training set for learning classifiers.