Figure 16:

Proportion of rejected test as a function of the significance level for pairwise equal distribution tests between pairs of graphs representing scans of the same subject (left) and different subjects (right). The black dotted line on the left figure corresponds to the identity, indicating the expected proportion of rejections assuming that the distribution of same-subject graphs is the same. The test constructed with MASE estimates either using the asymptotic distribution (MASE-A) or parametric bootstrap (MASE-B) to approximate the null is rejected more often than the OMNI test.