[Eeglablist] ICs with identical topographies

[Jason Palmer]

Dear Max, et al,

 

These two components are most likely slightly different-the thing to plot
would be the difference of the maps and of the activations (after reversing
the polarity/sign of one of the components-multiply the map and activation
by -1). The difference should not be exactly zero.

 

As Guillaume said, ICA can produce these "dependent subspaces" even if the
data is really full rank. These components will have mutual information with
each other, but should have close to 0 MI with other components.

 

Also, the time courses might be slightly time delayed, resulting in
instantaneous independence, but delayed dependence. A movie of the
backprojected sum of the maps might indicate a dynamic "wiggle" in the map
(potential distribution), which would result from the dynamic (non-static
spatial potential distribution) nature of the generating source.

 

Best,

Jason

 

 

From: eeglablist-bounces at sccn.ucsd.edu
[mailto:eeglablist-bounces at sccn.ucsd.edu] On Behalf Of Maximilien Chaumon
Sent: Wednesday, August 24, 2011 2:19 AM
To: Guillaume Rousselet
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] ICs with identical topographies

 

Hi Guillaume,

Well, it makes sense to my intuitive understanding...
The two components don't cancel each other, but then sum up, more or less.
Does that mean that I could somehow treat them as "one" component? 
I get your point with this superclean dataset example. And it also makes
sense with the other answers I got, with the rank of my data being not
exactly as high as I expect it to be. Now I need to understand why/how some
of my channels turn out to be linearly related...

Thanks,
Mx

2011/8/24 Guillaume Rousselet <Guillaume.Rousselet at psy.gla.ac.uk>

Hey Max,

 

your components don't cancel one another. In your example, the topographies
and the time courses have opposite signs, so if you multiply one by the
other, your two components are essentially the same.

ICA returns maximally independent components, and therefore can still be
correlated. The correlation can be quantified using the mutual information
plugin, which I use to confirm my own judgement about component similarity
when I look for eye blink ICs.

Imagine you have an absolutely clean signal, with only highly reliable
evoked activity and no noise at all. In that case you would expect to get n
times your number of electrodes the same component. In my experience,
cleaner datasets tend to have more correlated - almost identical - ICs.

Makes sense?

 

Best,

 

Guillaume


****************************************************************************
********
Guillaume A. Rousselet, Ph.D., senior lecturer & deputy post-graduate
convenor

Centre for Cognitive Neuroimaging (CCNi)
Institute of Neuroscience and Psychology 

College of Medical, Veterinary and Life Sciences

University of Glasgow
58 Hillhead Street
G12 8QB

 

http://www.psy.gla.ac.uk/staff/index.php?id=GAR01

 

Email: Guillaume.Rousselet at glasgow.ac.uk

Fax. +44 (0)141 330 4606 <tel:%2B44%20%280%29141%20330%204606> 

Tel. +44 (0)141 330 <tel:%2B44%20%280%29141%20330%C2%A06652>  6652

Cell +44 (0)791 779 7833 <tel:%2B44%20%280%29791%20779%207833> 

 

The University of Glasgow, charity number SC004401

****************************************************************************
********

 

On 23 Aug 2011, at 15:24, Maximilien Chaumon wrote:

 

Hi eeglabbers,

I sometimes get ICs with extremely similar topographies and time courses,
like on this slide <http://oszilla.hgs.hu-berlin.de/public/Similar_ICs.PNG>
.
I know that ICA returns independent components.
Does that not mean that they should not look the same?
I know the components are independent in a statistical sense, which is not
the same as uncorrelated, but still. I'm a bit surprised. What do these two
components mean if they cancel one another? well, do they?

Sorry if my question is naive, but what is happening?

The data is here <http://oszilla.hgs.hu-berlin.de/public/Similar_ICs.mat> .

Best,
Max

_______________________________________________
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

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110824/a707543c/attachment.html>