I have noticed that this preprint by Eyre, Fitzgibbon, et al. relies on 30 independent components. When I try to replicate a similar analysis, I find some RSNs where each lateral part is in a different component (visual, and visual association, for example). Is there a statistically correct manner of stitching these components together, or should I retry the group ICA with fewer components?
I realise this is a general rs-fMRI question and not uniquely concerning the dHCP dataset, but unfortunately I haven’t been able to find related responses in the FSL mailing lists or forums.
Thank you in advance!