[FieldTrip] ICA problem

Joseph Dien jdien07 at mac.com
Thu Jul 7 03:27:46 CEST 2011


I've had this same problem with EEGlab's runICA implementation when the rank of the data was less than the number of variables (i.e., it's an ill-conditioned matrix, which is to say it has multicolinearity).  Including the reference channel can do this (since it's just zeroes).  Using PCA to reduce the data to a subspace can help as well since the rank will now equal the number of variables (namely the number of factor scores will equal the number of retained factors).  Before you go the PCA route, it's worth trying to figure out why the data is ill-conditioned.  There is likely to be something wrong with one of your channels.  Maybe two of them are shorted together or something along those lines.  Or if you used mean mastoid reference and included both mastoids, that can do it too (since they will be exactly inversely correlated, one of them is statistically redundant, reducing the rank by one).  My EP Toolkit artifact correction routine (http://sourceforge.net/projects/erppcatoolkit/) takes all these sorts of things into account when performing eyeblink correction.

Cheers!

Joe


On Jul 6, 2011, at 8:44 PM, Sheraz Khan wrote:

> Hi Sylvana,
> 
> Is the data SSS (maxfilter), I normally have this problem when data is SSS, try doing ICA with PCA (set at 64 components).
> 
> Sheraz Khan
> Martinos Center
> Boston, USA
> 
> On Wed, Jul 6, 2011 at 2:09 PM, Stephen Whitmarsh <stephen.whitmarsh at gmail.com> wrote:
> Dear Sylvana,
> 
> >From the top op my head: When running the ICA, try only computing it
> for the MEG or EEG channels. (cfg.channel = 'MEG';), so it doesn't run
> on reference sensors in case of MEG.
> I have hunch that might help out.
> 
> Cheers,
> Stephen
> 
> 
> On 6 July 2011 19:37, Sylvana <sylvana.schister at utah.edu> wrote:
> > Hi all,
> >
> > I am trying to use ICA to clean EOG and ECG artifacts of neuromag data.
> > I am running ft_topoplotIC(cfg, comp), where 'comp' is the output of
> > ft_componentanalysis.
> >
> > I get the following error message:
> >
> >
> > ??? Error using ==> surf at 74
> > X, Y, Z, and C cannot be complex.
> >
> > Error in ==> ft_plot_topo at 184
> >  h = surf(Xi-deltax/2,Yi-deltay/2,zeros(size(Zi)), Zi, 'EdgeColor',
> > 'none', 'FaceColor', shading);
> >
> > Error in ==> ft_topoplotER at 753
> > ft_plot_topo(chanX,chanY,datavector,'interpmethod',cfg.interpolation,...
> >
> > Error in ==> ft_topoplotIC at 116
> >    ft_topoplotER(cfg, varargin{:});
> >
> >
> > The complex variable is 'Zi', which is taken from 'comp'. I can modify
> > the function to take the real part of the variable (or the absolute
> > value), but I am not sure if this would be the correct thing to do. In
> > case this is a bug, I thought I should report it. Any comments?
> >
> > Thanks for your help,
> >
> > Sylvana Schister
> > Univ. of Utah
> > Dept. of Bioengineering
> >
> > _______________________________________________
> > fieldtrip mailing list
> > fieldtrip at donders.ru.nl
> > http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> >
> 
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> 
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip


--------------------------------------------------------------------------------

Joseph Dien,
Senior Research Scientist
University of Maryland 

E-mail: jdien07 at mac.com
Phone: 301-226-8848
Fax: 301-226-8811
http://homepage.mac.com/jdien07/








-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20110706/ecd5b167/attachment-0002.html>


More information about the fieldtrip mailing list