Figure 4

Comparison of the best performing gene selection procedures identified by FiGS with other gene selection approaches in terms of the classification accuracy. The names of the compared gene selection procedures are abbreviated as follows: ttest, t-test; Wilcoxon, Wilcoxon rank sum test; and InfoGain, information gain method. 200 is the number of genes to select. varSelRF_SE0 and varSelRF_SE1 are two versions of varSelRF each with the standard error (SE) term set to 0 and 1, respectively. FiGS_best is the best gene selection procedure identified by FiGS; it produces the best classification accuracy with the smallest number of genes. The classification accuracy represented in the y-axis is 1-.632+bootstrap error.