<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Dear Paul,<br>
<blockquote cite="mid001e01c6f9fd$875a4500$0305a8c0@PAULDESKTOP"
type="cite">
<div><font face="Arial" size="2">
<div><font face="Arial" size="2">I am just starting to use EEGLab
5.03 and have a question out</font></div>
<div><font face="Arial" size="2">EEG.icaweights that </font>might
indicate a bug, but probably just indicates</div>
<div>my misunderstanding of the data structure.<font face="Arial"
size="2"> </font></div>
<div> </div>
<div><font face="Arial" size="2">My platform is Windows XP SP2
running </font><font face="Arial" size="2">Matlab 6.1.0.450.</font></div>
<div> </div>
<div>After running pop_runica using the interactive interface, I
exported</div>
<div>weights to a text file. The weights exported do not agree with </div>
<div>EEG.icaweights.</div>
<div> </div>
<div>I note, however, that if those exported weights are reloaded
into matlab as,</div>
<div>say, goo, then:</div>
<div> </div>
<div>EEG.icawinv * goo </div>
<div>gives the identity matrix</div>
<div> </div>
<div>however:</div>
<div> </div>
<div>EEGicawinv * EEG.icaweights</div>
<div>does NOT give the identity matrix</div>
<div> </div>
<div>is EEG.icaweights NOT supposed to be the weight (unmixing)
matrix?</div>
</font></div>
</blockquote>
This is because the exported "weight" matrix is <font><font
face="Arial" size="2">equal to EEG.icaweights*</font></font><font><font
face="Arial" size="2">EEG.icasphere. The sphere matrix is a matrix for
decorelating the data prior to running ICA. For more information see
the part on whitening the data at<br>
<a class="moz-txt-link-freetext" href="http://www.sccn.ucsd.edu/~arno/indexica.html">http://www.sccn.ucsd.edu/~arno/indexica.html</a><br>
<br>
If you type EEG.icawinv*EEG.icaweights*EEG.icasphere<br>
<br>
you will indeed get the identity matrix.<br>
<br>
Hope this helps,<br>
<br>
Arno<br>
<br>
ps: I have forwarded this message to the EEGLAB list as I thought it
might be of interest to other users.<br>
</font></font>
</body>
</html>