<P>Hello Masato,</P>
<P>To answer your question the Infomax algorithm, to my understanding, orders the data so that the first component holds the most amount of information about the original data set, with each successive component holding less information. This is done by analyzing the degree of variance of each component - the greater the variance the more information the component holds. FastICA for instance, will not do this, but rather order the components randomly.</P>
<P>I also have an inquiry of my own. I am wary of the results I am getting from running ICA on my data due to the composition of the components. When I choose Plot > Component Properties and examine the properties of a component, in the graph of magnitude (dB) vs. frequency (Hz) I get a distinct sinusoidal waveform. The waveform has a frequency of just over a 1 wave per Hz and an amplitude of about 15 dBells and extends all the way to 50Hz+. I noticed that in the EEGLAB tutorial PDF that on pg 61, the example graphs look like nothing of my own as the graph is fairly regular. Has anyone experienced this and does anyone know why the components contain signal in the 50Hz+ range?</P>
<P>Note: I am using 200Hz sampling rate on humans - interictal epilepsy patient data - with a 21 electrode montage.</P>
<P>Thanks,</P>
<P>Inuk Zand</P>
<P>London Health Sciences Centre <BR>Clinical Neurological Sciences <BR>University Campus, Rm B10-130 <BR>339 Windermere Road <BR>London, Ontario, Canada <BR>N6A 5A5 <BR>(519) 685-8300 ext. 35239</P>
<P><BR>----- Original Message -----<BR>From: Masato Nakazawa <MALISZT@UNM.EDU><BR>Date: Friday, July 29, 2005 7:09 pm<BR>Subject: [Eeglablist] Order in which independent components are presented<BR><BR>> Hello,<BR>> <BR>> I'm wondering in what order InfoMax and SOBI compoments <BR>> are computed. I mean, like the principle component, the <BR>> first SOBI (or infoMax) component accounts for the max <BR>> variance, and the second component second most variance? <BR>> or somehting else?<BR>> <BR>> I appreciate your answer,<BR>> <BR>> Masato Nakazawa<BR>> _______________________________________________<BR>> eeglablist mailing list eeglablist@sccn.ucsd.edu<BR>> http://sccn.ucsd.edu/mailman/listinfo/eeglablist<BR>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html<BR>> To unsubscribe, send an empty email to eeglablist-<BR>> <A href="mailto:unsubscribe@sccn.ucsd.edu">unsubscribe@sccn.ucsd.edu</A></P>