ROC and Type I error rate curves for the approaches utilizing a help model compared to the 2D χ
2 vs DW test. The structure and interpretation of the plots are the same as for Figure5: (A,C) are ROC curves, (B,D) are type I error curves, (A,B) are for the static, and (C,D) for the dynamic example. As can be seen, all methods are consistent, or weakly conservative. The ROC curves show that the 2D χ
2 vs χ
2 test (purple circles) is better than the 2D χ
2 vs DW test (green squares), and that the bootstrapped LHR (orange triangles), is best of all. In contrast LHP (pink diamonds) is completely uninformative, as it lies along the identity line.