[Eeglablist] Running a Second ICA

[Julie Onton]

David,

You can find out how many 'dimensions' your data contains (and therefore how many ICA components
you can realistically expect) by typing:

>> rank(EEG.data)

If the number returned is less than the number of channels, then you will need to run PCA on the
data before ICA.

Subjects need not have the same number of ICs (as they need not have the same number of channels
either). Even if you start with these two factors equal, different subjects are bound to have a
different numbers of 'dipolar' or brain-like components. If your strategy is rather to remove
artifact components and back-project to channels, the answer is still the same because you can't
expect all subjects to have the same type and amount of artifact activity. As long as you are
confident that what you are removing is identifiable artifact, I think it is valid to remove
different numbers of these components across subjects.

Hope this helps,  Julie

-- 
Julie Onton, PhD
Swartz Center for Computational Neuroscience
Institute for Neural Computation
University of California, San Diego
(858) 458-1927 ext 17
http://sccn.ucsd.edu/~julie

>
> Hi Everyone,
>
> I'm working on an analysis technique which requires that I run 2 ICAs
> on the same data set. The first is to remove standard eye artifact
> components, the second to remove components related to other
> artifacts. I don't use exactly the same data but for the second ICA I
> use about one third of each trial of the data I use for the first ICA.
>
> After listening to the discussions posted here I was wondering whether
> I should also run a PCA before the second ICA or whether I would not
> need to do that because I am not using exactly the same data for the
> second ICA. If I do need to run a PCA, should I reduce the data to
> the same number of principal components for each subject or could the
> number of principal components be equal to the number of ICs retained
> after the first ICA and therefore different for each subject?
>
> Many Thanks ...
>
>
> David Schwartzman
> Oxford Brookes University
>
>
>
> _________________________________________________________________
> Get Windows Live Messenger on your Mobile
> http://clk.atdmt.com/UKM/go/msnnkmgl0010000001ukm/direct/01/_______________________________________________
> 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