The dotted line to the right of the graph is SCheck’s cut-off value14. Student pairs whose Zb values fall beyond the cut-off are suspect; the response-independence hypothesis is rejected for these pairs, pointing to possible cheating.
In the case of the test center A data set, there are two Q-Q “blips” to the right of the cut-off. It has to be noted that each blip may actually represent more than one student pair; the precision of the graph is such as to preclude plotting each and every actual Zb value, much as was the case for the histogram of g2 test values seen earlier. However, SCheck’s graph makes it possible to see how outlying the outliers are, and it’s also possible to effectively gauge what’s happening to the immediate left of the cut-off. In the graph above, we see another pair of blips which are on the right, and set off from the main body. When I relaxed SCheck’s Type I error rate control for the test center A data set, the student pairs corresponding to these blips then came to fall to the right of the new cut-off Zb value15.