[FieldTrip] problem with ICA

Joseph Dien jdien07 at mac.com
Fri Oct 5 20:39:31 CEST 2012


One typically gets the complex numbers from ICA when the data have less than full rank.  For example, if there is a channel that is a flat line (maybe a bad channel or a reference) or when two channels are perfectly correlated (as when the data is mean mastoid referenced and one has included both mastoid channels as they will have a perfect -1 correlation).

Cheers!

Joe



On Oct 5, 2012, at 12:30 PM, Valentina Niccolai <vale.niccolai at gmail.com> wrote:

> Hallo.
> 
> I have a problem with the ICA, which I did not have before with the same dataset. It seems not to work properly because it gives complex numbers. The problem occurs also if I use older ft versions. I give in the ICA the output of ft_rejectvisual (already preprocessed data), which looks normal to me.
> 
> Here below is the output of ft_componentanalysis and in attachment the strange output when calling ft_databrowser: some or all the components are “dead” and the head views as well as the component number on the top disappear.
> 
> I am thankful for any help.
> 
>  
> Valentina Niccolai
> 
>  
>     cfg.method       = 'runica';   
> 
>     cfg.channel      = {'all','-EOG','-EMG'};
> 
>     comp = ft_componentanalysis(cfg, rej_vis_res);
> 
>  
>  
> The result of my component analysis looks like this:
> 
> comp =
> 
>  
>       fsample: 600
> 
>          time: {1x237 cell}
> 
>         trial: {1x237 cell}
> 
>          topo: [306x2 double]
> 
>      unmixing: [2x306 double]
> 
>         label: {2x1 cell}
> 
>     topolabel: {306x1 cell}
> 
>          grad: [1x1 struct]
> 
>     trialinfo: [237x1 double]
> 
>           cfg: [1x1 struct]
> 
>  
> and in comp.trial I get imaginary numbers:
> 
>  
> >> comp.trial{1}
> 
> ans =
> 
>    1.0e-18 *
> 
>   Columns 1 through 6
> 
>    0.0127 - 0.0000i  -0.0339 + 0.0000i  -0.0203 + 0.0000i  -0.0269 - 0.0000i   0.0176 + 0.0000i  -0.0538 + 0.0000i
> 
>   -0.0069 - 0.0000i   0.0075 - 0.0000i   0.0097 - 0.0000i   0.0033 - 0.0000i   0.0214 - 0.0001i  -0.0031 - 0.0001i
> 
>  
> <ica.jpg>_______________________________________________
> 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://joedien.com//










-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20121005/ab046029/attachment-0005.html>


More information about the fieldtrip mailing list