[Eeglablist] rejecting epochs vs rejecting continuous data

[Tarik S Bel-Bahar]

Thanks Francesca. OK!  Some extra notes below, best wishes.

************************************************************************

Whether rejecting entire epochs or smaller periods is up to the researcher
& their beliefs/biases/tendencies.
In your case, the best recommendation is to follow procedures from
published articles in high-quality journals using data similar to yours.
Remember ICA in eeglab does not care whether you feed it epochs or
continuous data, it mixes up all the timepoints you give it anyway.
The better question if whether you are feeding ICA the appropriate time
periods, and enough of them.

Note that once you have an ICA decompisition, you can do the following
which would minimize how many epochs you drop.
A.Clean your data just taking out the worst periods, as per eeglab
tutorials, but without epoching. Use only periods where the experiment was
actually happening.
B. apply the ICA to the continuous original data
C. remove bad/artifactual ICs (for example, eye-related and muscle-related
ICs)
D. then epoch
E. Then review the data and drop worst epochs

If you epoch from the start of your preprocessing pipeline, then yes, you
will be removing more than less of the time points of interest, because one
can only select full epochs.

An Alternative to consider here is to
A. Just take from the beginning of your experimental period to the end of
it.
B. Cut into small epochs (.5 or 1 sec)
C. Remove worst of these (fake) epochs
D. Run ICA
E. apply the ICA to the continuous original data
F. remove bad/artifactual ICs (for example, eye-related and muscle-related
ICs)
G. then epoch
H. Then review the data and drop worst epochs


It is however important to remember that you want to feed into ICA only the
periods of interest.
That is one reason to use epochs, as this will make sure that you only
feeding into ICA experimental data, and not data where the person is
waiting, doing nothing, or uninteresting "between epoch" pauses.

************************************************************************
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20170617/60957219/attachment.html>