[Eeglablist] reduce alpha distortions in event-related potentials

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Fri Jan 31 10:49:41 PST 2014


Dear Bastien,

I scanned the attached proceeding but I could not find how PARAFAC works. I
noticed that it uses wavelet transform for preprocess only for PARAFAC,
which would be certainly effective if you want to separate theta from alpha
in the end. I want to see why they used SOBI, how they evaluated
decomposition quality, how they identify alpha component etc, but the
authors did not explain it probably because the page limit of the
proceeding. So, honestly speaking I'm not fully convinced with the author's
claim.

If you have clear alpha in the data that 'contaminates' other ERP
component, that's a happy present if you are an ICA user since it promises
clear decomposition. Why don't you try EEGLAB's default ICA (extended
informax, which is better in decomposition performance; for detail see
Delorme et al.., 2012 PLoS One) on a few data and see the results. That
should not hurt!

Makoto






2014-01-30 Bastien Boutonnet <bastien.b1 at gmail.com>:

> Dear all
>
> (Thanks for the previous help regarding my STUDY statistics questions, esp
> Makoto & Arno -I have been busy with other things but will try your
> suggestions and see if I can move forward with that in a few weeks. Just
> thought I'd say thanks anyway)
>
> Someone in my lab finished collecting data destined to be analysed through
> classical ERP averaging. We looked at the data this morning (grand averages
> from 16+ participants) and it looks like the datasets are suffering from
> serious alpha contamination. Strong alpha contaminations are present in
> pretty much all participants.
>
> Is there any known way and possibly implemented way to remove the source
> of the alpha signals which are distorting the data? I thought of ICA
> immediately but a paper by Vanderperren (
> ftp://ftp.esat.kuleuven.be/sista/kvanderp/reports/MBEC_655.pdf) seems to
> claim ICA not to be very good for that and advocates a Parallel Factor
> Analysis. Has anything like that been implemented?
>
> If ICA was to be used have you got any suggestions as to how to go about
> it (I.e. how to identify good candidates once ICA decomp is done)
>
> Cheers,
> Bastien
>
> --
> Bastien Boutonnet, Ph.D.
> School of Psychology,
> Bangor University
> bastienboutonnet.com
>
> _______________________________________________
> 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
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20140131/3e0f27f8/attachment.html 


More information about the eeglablist mailing list