[Eeglablist] Comparing Components Across Trials

[Elliot Miles Asher Cohen]

Hello Eeglabers,

I am using eeglab as a preprocessing method to a BCI algorithm.  One
challenge in doing this is ensuring that the "same" components across
different trials get fed to the BCI algorithm.  Currently I am been
using eeglab's internal function matcorr to compare different trials
and then reorder the components according to the return value from
that function.  This seems to be giving me reasonable results, however
I was wondering if anyone else has looked into alternative methods of
comparing sets of components?

Does anyone know of any place where this issue has been discussed
either specifically in reference to eeglab or to ICA in general?  Or,
has anyone looked themselves into using any other clustering
algorithms such as k-means or EM?

Thank you very much for your time and help,

Elliot Cohen
emcohen3 at berkeley.edu
UC Berkeley