<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; ">Dear Martin and Simon,<div><br></div><div>I would be careful about pre-processing using PCA. I was just reviewing a PhD thesis (<a href="http://infosci.otago.ac.nz/carl-leichter-phd">http://infosci.otago.ac.nz/carl-leichter-phd</a>) that showed using simulations and real data that removing PCA components creates artifacts in the data, especially in the power spectrum. The same problem would arise if you run ICA and preprocess using PCA. I would be especially careful if you plan on removing these components from the original data, and then analyze the EEG power spectrum.<div><br></div><div>Arno</div><div><br></div><div><div>On 16 May 2013, at 15:25, Simon-Shlomo Poil wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div>Dear Martin,<br><br>As Makoto says, you make the channels less independent of each other.<br>It might be resonable* to reduce using PCA. One way to determine the<br>number of relevant dimesions could be<br>[COEFF, SCORE, LATENT] = princomp(EEG.data');<br>tmp = cumsum(LATENT);<br>nr=find(tmp/tmp(end)>0.975,1);<br><br>, which gives you the number of principle components explaining 97.5 %<br>of the variance.<br><br>*you can find previous mails discussing pro-/con- of PCA reduction on<br>the this list (I remember there was even a paper in prep? I didn't see<br>it come out)<br><br>Best wishes<br>Simon<br><br>--<br>Simon-Shlomo Poil, Dr.<br><br>2013/5/16 Makoto Miyakoshi <<a href="mailto:mmiyakoshi@ucsd.edu">mmiyakoshi@ucsd.edu</a>>:<br><blockquote type="cite">Dear Martin,<br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">If you apply a band-pass filter, your channel data become less independent<br></blockquote><blockquote type="cite">of each other i.e. rank-reduced.<br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">Imagine you apply an extreme band-pass filter, say 10-11Hz. All of your<br></blockquote><blockquote type="cite">channel data look very much like each other.<br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">Makoto<br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">2013/5/16 Krebber, Martin <<a href="mailto:martin.krebber@charite.de">martin.krebber@charite.de</a>><br></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Hi all,<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">I am currently working on an analysis were I split the data into low and<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">high frequency portions using a lowpass (cutoff 35 Hz) and a highpass<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">(20 Hz) filter, respectively. The idea behind this approach is to do the<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">ICA artefact rejection seperately on low and high frequency data in<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">order to be better able to reject high frequency muscle artefacts and<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">obtain a clearer brain signal in the gamma range.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">My problem is that, especially with the highpass filtered data, ICA<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">takes a very long time (roughly 5-10 times the usual) and even then the<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">decomposition does not look very clean. I tried to reduce the<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">dimensionality of the data (from 128 to 96) by applying the PCA<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">parameter in pop_runica and it is way faster. Is it justified, or maybe<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">even recommended to reduce the data dimensionality after filtering out a<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">considerable portion of the signal? And if so, is there a rule of thumb<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">about how much to reduce the data dimensionality?<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Thanks for any suggestions!<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Regards,<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Martin<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">_______________________________________________<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">To unsubscribe, send an empty email to<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">For digest mode, send an email with the subject "set digest mime" to<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br></blockquote></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">--<br></blockquote><blockquote type="cite">Makoto Miyakoshi<br></blockquote><blockquote type="cite">Swartz Center for Computational Neuroscience<br></blockquote><blockquote type="cite">Institute for Neural Computation, University of California San Diego<br></blockquote><br><br><br>--<br>_______________________________________________<br>Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>To unsubscribe, send an empty email to <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>For digest mode, send an email with the subject "set digest mime" to <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br></div></blockquote></div><br></div></body></html>