Hi
The point there is that small differences in the data can lead to very different decompositions, e.g. in one data set a network might appear split across components while in another data set it is being identified within one spatial map.... it boils down to a violation of your premise as you are not guaranteed to compare things *for the same RSN* but at best only for *pretty much the same RSN*.
What I would do is to run group-ICA and dual regression, then use the intermediate output (dr_stage1...) time course matrices to extract the network-specific time course and run the comparison.
hth
Christian
On 24 Feb 2012, at 03:32, Anna-Clare Milazzo wrote:
> Hi there,
>
> I would like to compare power spectra for the same RSN across different individual ICA sessions. For example, I might run single subject ICA on two scans from two separate subjects resulting in a sensorimotor network component for each subject. Is it valid to compare the power spectra from these two components or is there something about single subject ICA that would not make this comparison valid? The online pdf tutorial lecture for melodic hints that this might not be the case. I understand the point about the correspondence problem of RSN's across different subjects but I am not clear about the point of "Different splittings sometimes caused by small changes in the data" and would be grateful if someone might be able to clarify.
>
> Best,
> Anna
|