Thank you Randu,<br><br>So, should I do my rereferencing step like this?<br>I first compute the average M of my data across sensors, then rereference the data using an average reference, then add the M data as a new sensor. Is this correct?<br>
<br> Av = mean(EEG.data);<br><br> EEG = pop_reref( EEG, []);<br><br> EEG.data(end+1,:) = Av(:);<br> EEG.nbchan = size(EEG.data,1);<br> EEG.chanlocs(end+1).label = 'AveRef';<br><br>I don't really understand how adding another channel that is the average of all others will increase the rank of the data. And the result of svd(reshape(EEG.data,EEG.nbchan,EEG.trials*EEG.pnts)) does not change (except adding one point at the end, very close to zero).<br>
<br>Sorry to bug you with this, but I just want to be sure I'm doing it right.<br><br>Thanks,<br>Max<br><br><br><br><br><br><div class="gmail_quote">2011/9/12 Radu Ranta <span dir="ltr"><<a href="mailto:radu.ranta@ensem.inpl-nancy.fr">radu.ranta@ensem.inpl-nancy.fr</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">Hi Max,<br>
<br>
As a complement of information to what Scott was saying, re-referencing<br>
to the average without keeping the reference (i.e. the average) makes<br>
the rank of your data even smaller (basically, the sum of all your<br>
channels except the reference will be 0, that is you can always write<br>
one of them as -(sum of all others)). You might want to take a look at<br>
this paper, which suggests a which is the "best" re-referencing before<br>
ICA: "EEG montage analysis in the Blind Source Separation<br>
framework" (Salido-Ruiz et al, Biomedical Signal Processing and Control<br>
6(1), 2011).<br>
<br>
Radu<br>
<br>
Le jeudi 08 septembre 2011 à 15:02 +0200, Maximilien Chaumon a écrit :<br>
<div><div></div><div class="h5">> I think I found at least part of the solution to my problem:<br>
> I keep the reference channel (in my case an average of all electrodes)<br>
> in the dataset. This seems to reduce the rank of the data (the svd<br>
> vector drops close to zero on the last value). I think I get an<br>
> intuition of why it does so, but how am I supposed to do?<br>
> Overall, I've checked, the subjects on which I removed one or more bad<br>
> channels (after rereferencing) are fine. Their ICA looks nice, and the<br>
> svd values do not drop close to zero on the last value.<br>
> The problem is for those for whom I did not remove any bad channel.<br>
> Their ICA shows the symptoms shown on my last email, below.<br>
><br>
> So my questions end up being:<br>
> Do I have to reference my data to one electrode and not include it in<br>
> the ICA? and if so, why would anyone use an average reference before<br>
> ICA? what is recommended? I can't find anything about referencing the<br>
> data in the tutorial.<br>
><br>
> Thanks,<br>
> Max<br>
><br>
> 2011/9/7 Maximilien Chaumon <<a href="mailto:maximilien.chaumon@gmail.com">maximilien.chaumon@gmail.com</a>><br>
> Hi eeglabbers,<br>
><br>
> I am still having issues with ICA returning extremely similar<br>
> (but not identical) topographies. The gui (although I run a<br>
> version updated a few days ago) does not popup any suggestion<br>
> to reduce the rank. I only get this warning (Warning: fixing<br>
> rank computation inconsistency (68 vs 69) most likely because<br>
> running under Linux 64-bit MatlabAttempting to convert data<br>
> matrix to double precision for more accurate ICA results.) I<br>
> still get 69 components in the end.<br>
><br>
> Here's some more info:<br>
> the rank of the data is 68. The svd drops abruptly close to<br>
> zero at the last value. I have 69 electrodes (64 heancap +<br>
> 3EOG+2mastoids). I guess there's a gel bridge somewhere.<br>
> Although correlations between all electrodes don't reach .95.<br>
> When I let the ICA run with default options, I still get these<br>
> two components (always 'P3 like' components, this was<br>
> reproduced in other subjects). Their frequency profiles are<br>
> too good to be true, with low noise and a peak at 10Hz,<br>
> another one around 20Hz, see the figure. I would leave them<br>
> alone if they were not spoiling all my data: As I remove the<br>
> components, when I click this 'singles' button, which shows me<br>
> the trial by trial time course. I get what is shown on the<br>
> right of the figure. High frequency bursts appearing every now<br>
> and then, usually at times where there is high variability<br>
> across channels.<br>
> Removing both components resolves the issue, but I loose a<br>
> rather important part of the data.<br>
> Here is the spectopo at 60Hz. There is a strong artefact here.<br>
> The two components show a high power at all frequencies.<br>
> How could frequencies that do not exist in the input be<br>
> created by the ICA? I filter my data, before ICA below 45Hz.<br>
><br>
> I tried running fastica, asking for 68 components, no such<br>
> artifact appears but the decomposition looks much less nice,<br>
> at least with the parameters I've used.<br>
><br>
> So in the end, my question is:<br>
> How can I run an ICA without trouble if the rank of the data<br>
> is not equal to the number of electrodes? How can I identify<br>
> potentially gel bridged electrodes?<br>
><br>
> Many thanks,<br>
> Max<br>
><br>
><br>
><br>
><br>
> 2011/8/27 Arnaud Delorme <<a href="mailto:arno@ucsd.edu">arno@ucsd.edu</a>><br>
> Regarding the matrix rank, we have recently realized<br>
> that the rank function (and other rank function we had<br>
> programmed) are not fully reliable which is probably<br>
> with Max observes the component he observes. The<br>
> runica function should automatically decrease the rank<br>
> of the input data matrix. However, sometimes it does<br>
> not use the correct rank. We have modified the runica<br>
> GUI so that if the matrix is not full rank, it now<br>
> pops up a new window suggesting to the user a rank<br>
> reduced value. This value may be adjusted by the user<br>
> based on prior knowledge. For instance, if you have<br>
> removed 5 components from the data, you would reduce<br>
> the rank by 5 (and overwrite the rank that is<br>
> automatically detected if it is not correct).<br>
><br>
><br>
> Arno<br>
><br>
><br>
> On Aug 23, 2011, at 10:11 PM, John J.B. Allen wrote:<br>
><br>
> > Max<br>
> ><br>
> ><br>
> > I have observed that when the data are not full<br>
> > rank. You can test the rank of your data by<br>
> > reshaping your epoched data to a 2D matrix, and<br>
> > running the rank command, like this:<br>
> ><br>
> ><br>
> > rank(reshape(EEG.data,EEG.nbchan,EEG.trials*EEG.pnts))<br>
> ><br>
> ><br>
> > When I did this, your rank is 63, but you have 69<br>
> > channels, indicating that some channels are linearly<br>
> > dependent on others. I think this is the source of<br>
> > your problem, and if you remove those channels<br>
> > before running ICA, you should no longer see this<br>
> > issue.<br>
> ><br>
> ><br>
> > Best<br>
> ><br>
> ><br>
> > John<br>
> ><br>
> ><br>
> ><br>
> ><br>
> ><br>
> ><br>
> > On Tue, Aug 23, 2011 at 07:24, Maximilien Chaumon<br>
> > <<a href="mailto:maximilien.chaumon@gmail.com">maximilien.chaumon@gmail.com</a>> wrote:<br>
> > Hi eeglabbers,<br>
> ><br>
> > I sometimes get ICs with extremely similar<br>
> > topographies and time courses, like on this<br>
> > slide.<br>
> > I know that ICA returns independent<br>
> > components.<br>
> > Does that not mean that they should not look<br>
> > the same?<br>
> > I know the components are independent in a<br>
> > statistical sense, which is not the same as<br>
> > uncorrelated, but still. I'm a bit<br>
> > surprised. What do these two components mean<br>
> > if they cancel one another? well, do they?<br>
> ><br>
> > Sorry if my question is naive, but what is<br>
> > happening?<br>
> ><br>
> > The data is here.<br>
> ><br>
> > Best,<br>
> > Max<br>
> ><br>
> > _______________________________________________<br>
> > Eeglablist page:<br>
> > <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
> > To unsubscribe, send an empty email to<br>
> > <a href="mailto:eeglablist-unsubscribe@sccn.ucsd.edu">eeglablist-unsubscribe@sccn.ucsd.edu</a><br>
> > For digest mode, send an email with the<br>
> > subject "set digest mime" to<br>
> > <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br>
> ><br>
> ><br>
> > _______________________________________________<br>
> > Eeglablist page:<br>
> > <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">http://sccn.ucsd.edu/eeglab/eeglabmail.html</a><br>
> > To unsubscribe, send an empty email to<br>
> > <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<br>
> > digest mime" to <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br>
><br>
><br>
><br>
><br>
><br>
> _______________________________________________<br>
> Eeglablist page: <a href="http://sccn.ucsd.edu/eeglab/eeglabmail.html" target="_blank">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>
<br>
</div></div>--<br>
______________________________________________________________________<br>
E-mail : <a href="mailto:Radu.Ranta@ensem.inpl-nancy.fr">Radu.Ranta@ensem.inpl-nancy.fr</a><br>
<font color="#888888"><br>
Radu RANTA<br>
Nancy Université/INPL - ENSEM - CRAN<br>
2, Avenue de la Forêt de Haye Tel : <a href="tel:%2B33.%280%293.83.59.57.09" value="+33383595709">+33.(0)3.83.59.57.09</a><br>
F-54516 VANDOEUVRE-LES-NANCY Fax : <a href="tel:%2B33.%280%293.83.59.56.44" value="+33383595644">+33.(0)3.83.59.56.44</a><br>
FRANCE<br>
______________________________________________________________________<br>
<br>
<br>
</font></blockquote></div><br>