<p>I would try to err on the side of caution, and the paranoia is healthy till you have a defendable analysis path that makes sense to you. it seems your data is relatively clean, thanks for all the details, which are very useful.</p>
<p>If you have got channels to spare in a dense array it makes sense to drop some of the weirdest ones. </p>
<p>Ica definitely does a usually good job at parsing muscle and high frequency noise into relatively unique ics. Using templates representing muscle ic patterns is also possible as a cleaning method.</p>
<p>Ica has less indigestion when there are less of the most noisy channels, as well as the bad, disconnected, dangling, popping, and flat channels. I tend to remove channels dominated by exogenous artifactual noise.</p>
<p>For noise, consider tim mullen's CleanLine eeglab plugin, for wiping out artifact noise.</p>
<p>For your occipital channels, I would say it is cool to be draconian as long as you keep 60% of channels, that remaining channels are evenly dispersed, and no large groups of contiguous channels were removed from the scalp ROIs. </p>
<p>Not sure about decomposition destabilization due to high frequency noise, in some channels. </p>
<p>Your best bet is compare and ica with the noisy channels and without, after removing major artifactual time periods, of course.</p>
<p>It would be nice for eeglab to eventually<br>
Have an augmentation that allowed for catching a variety of types of bad channels. Note also number of bad channels per epoch is also an often used for bad channel rejection and epoch rejection criterion.<br><br><br><br>
<br></p>
<div class="gmail_quote">On Feb 18, 2012 12:53 PM, "Matthew Stief" <<a href="mailto:ms2272@cornell.edu">ms2272@cornell.edu</a>> wrote:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Sorry to barrage the list so much lately!<br><br>I was wondering if anyone would like to provide some more general guidance on channel rejection than has been discussed on the list in the past.<br><br>As has been noted several times before, the automatic channel rejection sometimes does not perform well, identifying acceptable electrodes and ignoring comically bad ones. Given the absence of comically bad ones, in most of my data it seems to identify electrodes that are not visibly any different from the others at least in scroll plot. For example: <a href="http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection1.jpg" target="_blank">http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection1.jpg</a><br>
<br>And just to take one channel identified in this way, it seems perfectly reasonable under channel properties as far as I can tell and no different from the adjacent one, here's a screenshot of that: <a href="http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection2.jpg" target="_blank">http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection2.jpg</a><br>
<br>Given this dearth of automatic guidance, I am left with the question of whether or not these channels are worth removing at all, and what are the hallmarks of a channel worth removing. Perhaps I'm just worrying too much and have admirable data, but I am too paranoid to think so. So, searching for ways in which to distinguish channels from one another at all, I experimented a little, and zooming out to a broader view it becomes clear that there are some bands of electrodes that have some extra high frequency noise in them that was not taken care of by the filter. You can see that here: <a href="http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection3.jpg" target="_blank">http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection3.jpg</a><br>
<br>Now though these channels have high frequency noise it's clear they've also got a lot of good neural information in them, and it seems to me that the ICA may be able to handle them admirably. Here is an example of one such channel: <a href="http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection4.jpg" target="_blank">http://s1153.photobucket.com/albums/p512/mstief/EEG%20Issues/?action=view¤t=channelrejection4.jpg</a><br>
<br>So first definite question: are such channels with high frequency noise likely to destabilize the ICA decomposition, especially if there are a lot of them, or will the noise be nicely separated out into a distinct component, leaving the underlying neural activity?<br>
<br>Second question: if these are the only visibly different channels I am dealing with, am I otherwise safe and can continue without rejecting any channels? I can always select parameters in automatic channel rejection to detect just a small handful of channels, but they never seem any different than the others so I'm reluctant to remove them.<br>
<br>Third question: If the general idea is that I should just use the automatic channel rejection function in some more refined way, is there still any guidelines on just what makes a channel bad, especially for the purposes of ICA? It may be helpful to know that I am not using ICA for artifact rejection but rather to isolate the visual P1. With that in mind should I be more or less draconian with occipital channels (i.e. do I want to be sure to save them to keep occipital activity, or remove them to avoid contaminating occipital ICs)?<br>
<br clear="all"><br>-- <br>_________________________________________________________________<br>Matthew Stief<br>Human Development | Sex & Gender Lab | Cornell University<br><a href="http://www.human.cornell.edu/HD/sexgender" target="_blank">http://www.human.cornell.edu/HD/sexgender</a><br>
<br><br>Heterosexuality isn't normal, it's just common.<br>-Dorothy Parker<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></blockquote></div>