Thanks Makoto,<br><br>I did not tweak the inside of the EEG structure, I just ran ICA a second time on the same EEG structure after removing the components. Perhaps it would be good to prohibit doing a second ICA if components have been removed from the dataset.<br>
<br>If I remember correctly I removed three components from the first decomposition, and found it funny that the second one then recovered components with extremely similar topographies, as shown <a href="http://oszilla.hgs.hu-berlin.de/public/2ICAs.png">here</a>.<br>
<br>I did update the wiki based on the beginning of this discussion and added the note you are referring to.<br><br>Max<br><br><br><div class="gmail_quote">2011/8/17 Makoto Miyakoshi <span dir="ltr"><<a href="mailto:mmiyakoshi@ucsd.edu">mmiyakoshi@ucsd.edu</a>></span><br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">Dear Max and Ronald,<br>
<br>
ICA randomizes all time points for every step of convergence, so it<br>
does not matter whether the data is epoched or continuous.<br>
<br>
Let me confirm this again: as it is described in the last line of the<br>
wikipage, 'components should NOT be rejected before the second ICA',<br>
if you want to improve the quality in decomposition. It is unusual<br>
that you obtain 15 ICs even after discarding 3 ICs out of 15... I<br>
don't think what you did is the supported method because once the data<br>
rank is reduced (e.g. 15 -> 12 as a result of discarding 3 ICs) it<br>
never recovers. I guess that probably you copied (ALL)EEG.data to the<br>
other (ALL)EEG.data directly which does not have ICA weights. This is<br>
not the way to improve quality in decomposition, so do not bother to<br>
test this method.<br>
<br>
Makoto<br>
<br>
2011/8/16 Maximilien Chaumon <<a href="mailto:maximilien.chaumon@gmail.com">maximilien.chaumon@gmail.com</a>>:<br>
<div><div></div><div class="h5">> Thanks Ronald,<br>
> I understand your point with the dimensions.<br>
> I thought ICA does not care about time, though? So concatenated, epoched or<br>
> not, the results should be the same, right?<br>
> Max<br>
><br>
> 2011/8/16 Ronald Phlypo <<a href="mailto:Ronald.Phlypo@ugent.be">Ronald.Phlypo@ugent.be</a>><br>
>><br>
>> Dear Max,<br>
>><br>
>> the problem lies in the maximally allowed number of components in both<br>
>> decompositions. The first decomposition may allow for as many sources as<br>
>> sensors (15 in your case). However, once 3 artefacts have been removed, your<br>
>> data dimension reduces to 12, which allows to estimate a maximum of 12<br>
>> sources only. To circumvent this problem, I suppose what the wiki means is<br>
>> to do local decompositions first (trial by trial or epoch by epoch) and then<br>
>> concatenating the trials/epochs again after their correction and before the<br>
>> second "joint" decomposition. Since in this case the artefact removal is<br>
>> nonlinear, it does not reduce the dimension of your concatenated data.<br>
>><br>
>> Hope this helps,<br>
>><br>
>> Ronald<br>
>><br>
>> PS: you might also want to have a look at<br>
>> <a href="http://www.hindawi.com/journals/cin/2007/075079/cta/" target="_blank">http://www.hindawi.com/journals/cin/2007/075079/cta/</a> where short time and<br>
>> long term windows are used jointly for artefact removal. The text refers to<br>
>> literature on the mean duration of electrophysiological processes to<br>
>> motivate this decision.<br>
>><br>
>> Le 15/08/2011 18:09, Maximilien Chaumon a écrit :<br>
>><br>
>> Hello all,<br>
>><br>
>> I'm currently cleaning data before working with components.<br>
>><br>
>> I cut my dataset into epochs<br>
>> reject epochs where signal is bad<br>
>> run an ICA<br>
>> find blink and muscle components, reject them<br>
>> run an ICA again, and look at my components.<br>
>><br>
>> ... as I understood was suggested at the bottom of this page<br>
>> <a href="http://sccn.ucsd.edu/wiki/Chapter_01:_Rejecting_Artifacts" target="_blank">http://sccn.ucsd.edu/wiki/Chapter_01:_Rejecting_Artifacts</a><br>
>><br>
>> Then the ICs look very nice but come in pairs of extremely similar<br>
>> topographies with different time courses, as shown on this picture.<br>
>> I am wondering what happened here. I can imagine that rejecting components<br>
>> before running the second ICA is what went wrong... But why did I read that<br>
>> on the wiki?<br>
>><br>
>> Thanks a lot for any advice.<br>
>> Max<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<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 digest mime" to<br>
>> <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><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<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 digest mime" to<br>
>> <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><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<br>
> <a href="mailto:eeglablist-request@sccn.ucsd.edu">eeglablist-request@sccn.ucsd.edu</a><br>
><br>
<br>
<br>
<br>
</div></div>--<br>
<div><div></div><div class="h5">Makoto Miyakoshi<br>
JSPS Postdoctral Fellow for Research Abroad<br>
Swartz Center for Computational Neuroscience<br>
Institute for Neural Computation, University of California San Diego<br>
</div></div></blockquote></div><br>