Figure 4

Empirical Distribution over Permutation Test on Performance of Feature Selection Method. Plots showing the (A) prediction performance difference between the feature selection results of L1SVM and that of t2-statistics (same number of selected features), and (B) stability of L1SVM. The x-axis is the statistic of interest of the corresponding permutation test. The y-axis is the frequency of the given value of the statistic of interest. The red dotted line indicates the observed statistic of interest and a blue bar describes the frequency at a given value of the statistic of interest from the permutation test.