Fig 9. Scatter of the upper-tail probabilities of the two tests in the presence of a repeated SSE.

The upper tail probability of the first test (y-axis) in relation to that of the second test (x-axis) for all models 0, …, 9. Each model is repeated 100 times and in each repetition an SSE composed of 7 links and 5 neurons/link is injected at two random points in time. The represented values max(1 − P_{ik, jk}) and max(1 − J_{ik, jk}) are taken each as the maximum (i.e. least significant) significance value across all entries ${\mathcal{S}}_{DS}:=\{(i_k,j_k),k=1,2,\dots ,7\}$ of the intersection matrix composing the embedded DS. The figure shows for each model the cloud of the 100 points (max(1 − J_{ik, jk}), max(1 − P_{ik, jk})), one per simulation. The red lines mark the significance levels 1 − α₁ (horizontal) and 1 − α₂ (vertical) for individual tail probability values 1 − P_ij and tail joint probability values 1 − J_ij, respectively.