[Eeglablist] ICA using PCA

[Makoto Miyakoshi]

Dear Alessandra,

I would recommend you go back to continuous data to run ICA since you have
as many as >200 channels.

307 x 200 = only 60,000 with which you can't get a good result. The rule of
thumb is that if you have one hundred channel data, you need one million
datapoints.

> We can't understand the formula (k/N^2=N stable components)

(235^2) x 30 = 1,656,750 datapoints you need to decompose 235 channel data.
Actually, they say it's not fixed to be 30 but it should increase as the
number of the channels increase from 30. So you'll probably need 2 to 3
million datapoints (which is nearly unrealistic, I know...)

Makoto

On Tue, Jan 5, 2016 at 5:47 AM, Alessandra Di Pietro <
dipietroalessandra at hotmail.it> wrote:

> Hi everyone,
>
> I have a question: we are recording 235 channels , sampling rate 256,
> Epoch start -0.2 (s), epoch end 1 (s), frames per epoch 307, epochs 200.
>
> We want to perform ICA using PCA. How many component we need? We can't
> understand the formula (k/N^2=N stable components)
>
> Thank you
>
> _______________________________________________
> 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/20160106/4abef40d/attachment.html>