[Eeglablist] group level analysis

[Scott Makeig]

Some comments below.  -Scott Makeig


Dénes Szucs wrote:
> Thank you for your earlier reply.
> 
> Another question:
> To assess group level data is the process outlined below the right 
> one?
> - perform ICA on individual data
> - identify relevant components (ICs) - ?  according to their relation to 
> stimulus onset and response time and similar power spectra?

Actually, different combinations of measures could be used - ERSPs, 
dipole locations, ERPs or ERP images, etc.

> - identify identical relevant components across subjects -- ? Is it 
> usually possible to identify components across subjects? Is there an 
> objective method for this or the researcher has to "subjectively" 
> judge the similarity of ICs?

See Makeig et al Science 2002 for an example of clustering components 
across subjects. Many components do fall into clusters - of course, the 
number of clusters may depend on how many subjects and components you 
are clustering, as well as the 'dipolarity' of the clustered components, 
the measures used, etc.

> - average identical component loadings across subjects
> --> is it possible this way to describe "group level" components?

Yes. See the Science paper at http://sccn.ucsd.edu/science2002.html

> But:
> - What if some components are missing in some persons? Sould 
> one accept this as an indication of individual variability and simply 
> use only components accepted as identical from other subjects? Or 
> is it due to data quality not good enough? 

I think your thinking here is correct - at least, to present knowledge. 
There may be several reasons why a component type would  not show up in 
every subjects - insufficient number of channels, data quality, number 
of time points, relative strength of the component activity, or absence 
of a similar EEG-producing synchronization in some some subjects...

> It it possible to tell the 
> quality of the data from the point of view of the ICA algorithm? (apart 
> from the No of data points needed)


We are working a measure of ICA decomposition quality based on dipole 
fitting - and hope to produce a manuscript soon.

> 
> thank you, denes
> 
> 
> 
> Denes Szucs
> Lecturer in Neuroscience and Education
> University of Cambridge, UK
> Shaftesbury Road, Cambridge, CB2 2BX, UK
> Tel: +44-(0)1223 369631
> Fax: +44-(0)1223 324421
> 
> _______________________________________________
> Eeglablist mailing list Eeglablist at sccn.ucsd.edu
> Eeglablist page: http://www.sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsub at sccn.ucsd.edu
> To switch to digest mode, send an empty email to eeglablist-digest at sccn.ucsd.edu