Hi Till,

Yes, definitely, there are of course a lot of good theoretical reasons to want to better understand transdiagnostic commonalities and differences, going beyond simpler patient-control comparisons. But even purely from a methodological perspective, focusing only on patient-HC differences has the inherent risk of overinterpreting any differences between the result across different patient groups (e.g. say, any differences between the "Panic-HC" results and the "Social Phobia-HC" results may or may not reflect true "Panic - Social Phobia" differences, and there is no way to know that other than directly testing that inter-patient contrast). In general, while differences between the patient groups can be directly tested using a standard ANOVA design (e.g. evaluating "any differences" between the groups with a [-1 1 0; 0 -1 1] contrast), or by using correlation analyses across meaningful behavioral dimensions, evaluating "commonalities" is always bit more ambiguous. One approach, for example, is to combine an F-test that looks at 'any effects' across the patient groups (i.e. a [1 0 0;0 1 0;0 0 1] contrast), and then use an exclusive intersection of those results with the previous "any differences" contrast in order to find region where there is a significant effect in at least one of the patient groups, and where at the same time there are no significant differences between the three patient groups. Of course another simpler/common approach is to evaluate common effects by averaging across the three groups (e.g. using a [1/3 1/3 1/3] contrast across the patient groups). Other approaches include data-driven clustering to try to identify similar sub-groups across all patients, evaluating then the association between the resulting clusters and the diagnostic categories, and similar strategies.

Hope this helps
Alfonso