Figure 12

Simulation II: Comparison between Pearson’s Chi-square test and SCBS procedure in feature selection. Four curves presented in the plot are based on the true positive rate (TPR) by two methods under two different causal effects β=1 and β=2. Sample sizes are set to be 500, 1000, 2000 and 5000. The total number of positives is restricted to be 49 for both methods and the TPR is calculated as the number of true positives divided by 49.