[Eeglablist] Different filters, different matrix ranks?

[Thomas Schüller]

Hi eeglab list gurus

I have a problem that I really can't figure out on my own. I have eeg 
data recorded from 64 electrodes + reference  (CPz).  I find that using 
different filters changes the rank of the matrix in ways I would not 
have anticipated.

First, I used a Butterworth highpass (.5Hz) and a Chebychew lowpass 
(25Hz) filter on the raw data and then reconstructed CPz in the process 
of re-referencingto average, and then epoched the file resulting in a 
EEG.data(:,:,1) matrix with rank 65.
When computing the ICA using runica() I get a warning: "fixing rank 
computation inconsitency (65 vs 64) most likely because running under 
Linux 64-bit Matlab".
The server that I ran it on is indeed 64-bit, but it's a windows 
server... the ICA creates 65 ICs, as many as there are Channels!

Second, I used the eegfilt function (highpass: .5 / lowpass:25) on the 
raw data and repeated the process as described above. Now, after 
reconstructing and re-referencing EEG.data(:,:,1) has a rank of 64.
When computing ICA i get the warning: "Data rank (64) is smaller than 
the number of channels (65)"
and sure enough ICA composes only 64 ICs!

Does any one know why this is happening?

Many thanks!

Thomas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110622/dcf2a6b5/attachment.html>