Dear experts,

I have 2 questions regarding graph theoretical anaylsis (in CONN) with connectivity measures resulting from different frequency bands:
I'm trying to investigate structural and functional differences between 2 groups (S, controls). Structurally, I found a cluster in the left frontal pole region with the VBM analysis for which I extracted this specific ROI (= VBM ROI) and used it as an additional ROI in CONN. This VBM ROI spatially overlaps with the left frontal pole ROI from the CONN default atlas (Oxford-Harvard Atlas).
I would like to find out whether this VBM ROI has more global efficiency in S compared to controls. However, during preprocessing I partioned the band-pass filtered time series of this VBM ROI (.008- .09) into 3 equally non overlapping smaller time series (Freq1, Freq2, Freq3).But I did this only for this ROI not for the other ones (all other ROIs have the typical frequency band .008 - .09)
Now my questions are the following:
1) Is it a problem that in the adjacency matrix I have two ROIs (VBM ROI and the default left frontal pole ROI) that spatially overlap? I could imagine that the connecitivty values are not independent from each other, are they?
2) Is it a problem for the graph theoretic analysis (even if I leave out the default left frontal pole ROI as node in the adjacency matrix and I include just one frequency band of my VBM ROI for the analysis) if the connectivity values for my VBM ROI result from a different frequency band (Freq1) than the other ROIs (0.008 - 0.09)?
I get a great result (global efficiency for VBM ROI: S>Controls) if I include Freq1 of my VBM ROI into the analysis but I don't get anything if I include the same frequency band for my VBM ROI as for all the other ROIs. I wonder if my result with for the VBM ROI and its Freq1 is merely due to the fact that it has a different frequency band than all the other ones?
I would be tremendously grateful if any of you experts could quickly give me some thoughts and suggestions on this.
Thanks a lot,
Maxi