[Eeglablist] How to perform cluster analysis

[Arnaud Delorme]

>
> In the article "Dynamic Brain Sources of Visual Evoked Responses" 
> (Makeig et al, 2002) there is cluster analysis in order to see how 
> components are common across subjects. This is a normal problem when 
> analyzing real data: to have a method to compare the components 
> obtained across subjects. Is there any EEGLAB function that can make 
> this  kind of analysis?

The simplest way to perform cluster analysis on two subjects is to use 
the EEGLAB matcorr() function (from the command line only) which compute 
the correlation between the scalp map of two subjects. Unfortunately 
this function does not take into account the activity of the two 
components.
To cluster several subject and combine spectral and topographical 
information, you may use the clusterdata() function from the matlab 
statistical toolbox (we also have our own function which will be 
included a future EEGLAB release). Cluster analysis of ICA components 
often ends up in a sub-optimal local minima and can lead to spurious 
results. We advise to first cluster ICA components manually on a couple 
of subjects before running automatic algorithms.

Arno



> Is there any paper that explains it? (I've read Enghoff but it would 
> be great to have another paper)?
>  
> Thank you. Yours sincerelly:
>  
> Josep Marco



-- 

*Arnaud Delorme, Ph.D.*
Computational Neurobiology Lab, Salk Institute
10010 North Torrey Pines Road
La Jolla, CA 92037 USA

*Tel* : /(+1)-858-458-1927 ext 15/
*Fax* : /(+1)-858-458-1847/
*Web page *: www.sccn.ucsd.edu/~arno <http://www.sccn.ucsd.edu/%7Earno>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20030714/34e8b0f2/attachment.htm