[Eeglablist] Weights matrix complex-valued after runica and NaN after binica.

[Joseph Dien]

One way this can happen is if the data has a rank less than the number of variables (that is, some of the variables are informationally redundant or uninformative).  For example, are any of the channels perfectly correlated?  This can happen if you use a mean mastoids reference and include both of the mastoid channels (they will have a perfect -1 correlation).  This can also happen if you have a bad channel and it is completely flat.  There are a number of other scenarios that can cause this too.

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

Joseph Dienhttp://homepage.mac.com/jdien07/
On Sep 06, 2011, at 02:18 PM, Paul Kieffaber <pdkieffaber at wm.edu> wrote:


EEGlabers,
I'm having a very strange issue running runica/binica from the command
line in 64-bit Matlab 7.12.0 (R2011a) under Ubuntu linux. The data
are 65 channels by 700 samples by 500 trials. Runica.m (from command
line or EEGlab menus) is returning a complex-valued "weights" matrix
and Binica.m (run from command line) is returning "weights" and
"sphere" matrices full of NaNs. I've run ICA on this computer before
with no trouble. This is happening with an entire sample of 30
participants. There doesn't seem to be anything unusual with the
data. Now the really bizarre thing... for some of the participant's
data, runica/binica each run successfully if I use only a subset of
the trials. In some cases they will only run (and return real
numbers) with just one trial, but other subjects will run with up to
50 trials. I haven't ever seen anything like this and am now
completely baffled.

Any help/advice would be greatly appreciated.

paul
-- 
-----------------------------------------------------
Paul Kieffaber, Ph.D.
Assistant Professor
Department of Psychology &
Program in Neuroscience
The College of William & Mary
(757) 221-1965
pdkieffaber at wm.edu
-----------------------------------------------------
_______________________________________________
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/20110907/e7086670/attachment.html>