[Eeglablist] PowPowCAT Group Analysis

[Makoto Miyakoshi]

Hello Edward,

Thank you for your interest in PowPowCAT.

There is a general difficulty in performing a post-ICA group-level analysis
with any factorial design due to Post-ICA Inter-individual Consistency
(PIC-IC) problem. Put it simply, there will be different number of ICs
representing different 'brain' activities across different locations.
Typically, we use statistical approach to address this issue, but as you
can imagine this is pretty complicated and messy. For example, you can no
longer have all the subjects at a selected location, and first of all you
cannot specify a priori spatial ROI because IC's localization does not fill
in the space uniformly but very sparse (typically 5-20 locations per
subject).

The current group-level solution for PowPowCAT is for ICA or ECoG purists.
You can design a probabilistic spatial framework to address this issue,
like I did in developing groupSIFT, but it is going to be pretty
complicated.

I can imagine a following solution (in fact, I remember I developed
something very similar for my collaborator Ilaria):
(1) You specify spatial ROI at MNI XYZ coordinate
(2) You determine effective radius from that point, say 20 mm
(3) For one group, find all the ICs within that radius to perform a
group-level analysis
(4) Do the same for the other group.
(5) Perform two-sample t-test on the result from (4)

I have the Ilaria special, the basic framework of which can be used for
your purpose. If you are interested in this solution, let me know.

The abovementioned ROI-based solution is actually an effective solution for
PIC-IC problem, not only for PowPowCAT but also for ERP, ERSP, PSD, ITC,
etc.. The prototype solutio for Ilaria showed surprisingly good results,
seemingly sensitively differentiating adjacent anatomical ROIs, at least
better than I expected (I'm a pessimist)

Makoto

On Thu, Dec 8, 2022 at 7:31 PM Edward Lannon <elannon at stanford.edu> wrote:

> Hello,
>
> I am working with clinical EEG resting state data and I am very interested
> in using PowPowCAT. Do you have any recommendations of how to used the
> PowPowCAT output data to test between group differences (Healthy CTL vs
> Clinical Population)?
>
> Thanks for the help,
>
> Edward
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>