[Eeglablist] practical questions about Dipfit

Bernhardsson Jens Jens.Bernhardsson at miun.se
Fri Nov 10 03:03:43 PST 2017


1.       After the rejections of noisy data in the time domain (noisy epochs) because you run the IC analysis after those exclusions but before the exclusion ICs. Fitting dipoles to all ICs will help you in classification of artifact based on residual variance and dipoles found outside the head volume.

2.       It depends on what you want to do. If you want to run time frequency analysis from 3 Hz you need long baseline and thus epochs. However, longer epochs gives you higher probability of having non frequent artifacts (jaw clinching, dry eye blinks, head movement etc.) which are hard for the ICA to decompose. If it is necessary, you can transfer ICs from ICA performed on shorter epochs to longer epochs.

3.       I believe dipfit() default is residual variance minimization, if you set a RV threshold to 100% local minima is always possible. After that it’s your decision what cut off to use, I believe recommended RV is 15% and to exclude dipoles found outside the head volume, this can easily be done when you use the STUDY function.


Från: eeglablist [mailto:eeglablist-bounces at sccn.ucsd.edu] För Carolina Pletti
Skickat: den 9 november 2017 16:44
Till: eeglablist at sccn.ucsd.edu
Ämne: [Eeglablist] practical questions about Dipfit

Dear EEGlab community,
I have a few questions regarding Dipfit.
1 - would you recommend fitting the dipoles before or after artifact rejection?
2 - Is there any suggestions on whether to run it on long (e.g. 3000 ms or more) vs short (e.g. 1000 ms) data epochs?
3 - what should one do when it finds a local minimum, or when a local minimum is possible? Is it normal that it is always possible?
Thank you very much!
Carolina Pletti
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20171110/3214b2d8/attachment.html>

More information about the eeglablist mailing list