Hi Chaleece,

Some thoughts on your questions below
Best
Alfonso
Yes, sorry about that, first by 'extract values' I really meant 'import values' and also you are right that the instructions were for seed-to-voxel analyses. You could do exactly the same for ROI-to-ROI analyses but in this case it might be even simpler, since those effects that you care about will already be displayed in the barplots on the main CONN window. When you select your target ROIs(s) (by either clicking on the corresponding ROIs in the brain display or selecting those ROIs in the 'analysis results' list) you will get three bars for each ROI displaying the controls, patients, and severity effects, respectively. You may then simply look directly in those plots at whether the sign/direction of the third bar (the severity effect) is the same or not as the difference between the first and second bars (the controls vs. patient effects) to figure out whether the severity changes are in the direction of increased or decreased differences in connectivity between your two groups (but see the note on point (3) below since that is relevant on how to properly interpret the difference between the first and second bar -between-group differences-)

Yes, exactly right, if severity decreases with higher SeverityScore numbers then you just want to look at the opposite signs. For example, a negative "SeverityScore" effect means that connectivity values are increasing with higher symptom severity (ie connectivity values increase with lower SeverityScores values), so those changes will tend to *increase* the patients-control differences if the patients also show higher connectivity values than controls, while the same changes will tend to *decrease* the patients-controls differences if the patients show lower connectivity values than controls.   

This is a very good question and it depends on exactly what you mean by the "difference in connectivity between patients and controls" in the presence of severity effects so let me elaborate on that a little bit before actually answering your question. Because you are looking at regions where connectivity varies with different severity levels, then the size of the difference in connectivity between patients and controls will change depending on the level of severity of the patients that you are looking at, so you need to precisely define at what level of severity you are evaluating your "patients" group when estimating the patients vs. control differences in connectivity. One typical approach is to consider the average severity level in your patients sample. Another common approach would be to consider an "ideal" level of severity in patients that would make it most comparable to the control group (for example, if there is a SeverityScore value that represents the absence of any symptoms then you might want to compare the connectivity in controls vs. the connectivity in patients at this "no symptoms" level for a different interpretation of the patients vs. control connectivity differences). The level of SeverityScores at which you want to evaluate the connectivity effects in your patients group is effectively controlled by centering (or not) of the SeverityScores covariate. If you wish to evaluate the connectivity in patients at the average SeverityScores level then center (remove the mean of) the SeverityScore second-level covariate before defining your second-level model, while if you wish to evaluate the connectivity in patients at some other arbitrary SeverityScores level then again simply center the SeverityScore covariate at the desired level (subtract the desired SeverityScore level from the SeverityScore values). In your case, since you mention that SeverityScore values decrease with higher severity it probably does not make too much sense to evaluate the connectivity values in patients at the zero-level of the SeverityScores variable, so it is probably best to center/demean those covariates (subtract the average values within the patient group). After you do that, you will notice that in the barplots of your second-level model the first and third bars (the 'controls' and 'severity' effects) remain exactly the same as before, but the second bar (the 'patients' effect) will change, now representing the average connectivity within the patients group (before it was representing the estimated connectivity within the patients group at the zero-level of the SeverityScores covariate). After all of this I can finally answer your original question, and in this case then yes, if you find a significant 'severity' effect but no significant 'patients-controls' differences, this indicates that while severity influences the connectivity in patients, that effect is not contributing to any difference in connectivity between controls and patients in your sample.

Hope this helps and let me know if you would like me to further clarify any of the above
Alfonso