[Eeglablist] Questions related to event-related (phase) coherence

[Makoto Miyakoshi]

Dear Jung,

> 1. While I appreciate that it's clearly explained in Delorme and Makeig
(2004) that ERCOH is a measure of the degree of synchronization between two
signals, I would like to know how it differs from other phase
synchronization measures, e.g. PLV. I couldn't find any review that
includes ERCOH in the comparison of synchronization measures. Does anyone
have a reading suggestion?

I remember there are more than enough variations and there are indeed
subtle differences from one another. However the basic idea is more or less
the same, and they are all valid (there was an interesting update about
inter-trial phase coherence a few years ago from Neuroscience Institute in
Cuba, check it out if you are interested in). You don't need to be nervous
about it. Just use what you found first.

> 2. When calculating ERCOH values for every time-frequency points to get
the time-frequency plot from a single subject's epoched data, does EEGLAB
calculate ERCOHs from single epochs then average, or calculate ERCOHs from
the averaged epoch data? My intuition is that phase information will be
somewhat lost (or blurred) when epochs are averaged, so EEGLAB should
calculate ERCOHs from individual epochs first and then average across
epochs. Does it?

Yes, you are right. If there is an average trial only, you can't compute
phase consistency across trials.

> 3. If PLV and ERCOH are comparable measures of phase synchronization,
they must use the same information from the signals (e.g. phase
information) to calculate the values. Then, if I decide to use PLV instead
of ERCOH, I believe it should be fairly easy to modify the code for ERCOH
(substitute ERCOH equation with PLV equation) to get a time-frequency plot
showing PLV values, and a matrix of PLV values. Is anyone familiar with
such procedure, and willing to provide advice on this?

In EEGLAB, newtimef() computes wavelet transform. My advice is that
newtimef() is one of three trickiest functions (other two are eegplot() and
erpimage(), of course in my opinion) so get ready. The concept seems easy,
yes I agree.

> 4. In time-frequency decomposition, how long is the default time steps of
the overlapping time window in EEGLAB? Is there a way to modify the time
steps? For example, I would like to use 3-cycle Morlet wavelet for the
decomposition, and I would like to have this window slide in time steps of
10ms.

See slide 21 of the following material
http://sccn.ucsd.edu/mediawiki/images/1/19/C2_A3_Time-frequencyDecAndAdvancedICAPracticum.pdf

> 5. For eyeblink reduction method, I understand that it's the most common
practice to use ICA. I also understand that there are particular pluses and
minuses of various artifact reduction methods. From my lit review, I
couldn't find any sort of preference for a method over others in studies
that looked at coherence, and couldn't find any review on this matter
either. Does the fact that ICA is a powerful tool justify using ICA for
artifact reduction in preprocessing data for coherence analysis? Is there
any issue with this method with particular regards to coherence analysis?

We claim that ICA is a good tool. We recommend you analyzed ICA-decomposed
EEGs, not channel EEGs cleaned with ICA; when you have unmixed source
activities, why do you want to mix them again??

> 6. When missing reference channels, would people prefer using a single
reference (i.e. Cz) or common average reference, in prepping data for later
coherence analysis?

Use common average reference.
There is good alternative which I have never tried.
http://www.neuro.uestc.edu.cn/rest/ This is supposed to be better than
average reference. Check also Nima's PREP paper for cleaning before average
referencing.
http://journal.frontiersin.org/article/10.3389/fninf.2015.00016/abstract

Makoto


On Thu, Dec 10, 2015 at 12:43 AM, Jung Hwa Han <junghwahan at gmail.com> wrote:

> Dear EEGLAB list,
>
> I am a naive Matlab user (not familiar with math either) who is trying to
> learn to use EEGLAB for analyses involving event-related band power change
> and (cross-channel) phase coherence. I'm working with Neuroscan .cnt files.
> Several questions regarding phase coherence:
>
> 1. While I appreciate that it's clearly explained in Delorme and Makeig
> (2004) that ERCOH is a measure of the degree of synchronization between two
> signals, I would like to know how it differs from other phase
> synchronization measures, e.g. PLV. I couldn't find any review that
> includes ERCOH in the comparison of synchronization measures. Does anyone
> have a reading suggestion?
>
> 2. When calculating ERCOH values for every time-frequency points to get
> the time-frequency plot from a single subject's epoched data, does EEGLAB
> calculate ERCOHs from single epochs then average, or calculate ERCOHs from
> the averaged epoch data? My intuition is that phase information will be
> somewhat lost (or blurred) when epochs are averaged, so EEGLAB should
> calculate ERCOHs from individual epochs first and then average across
> epochs. Does it?
>
> 3. If PLV and ERCOH are comparable measures of phase synchronization, they
> must use the same information from the signals (e.g. phase information) to
> calculate the values. Then, if I decide to use PLV instead of ERCOH, I
> believe it should be fairly easy to modify the code for ERCOH (substitute
> ERCOH equation with PLV equation) to get a time-frequency plot showing PLV
> values, and a matrix of PLV values. Is anyone familiar with such procedure,
> and willing to provide advice on this?
>
> 4. In time-frequency decomposition, how long is the default time steps of
> the overlapping time window in EEGLAB? Is there a way to modify the time
> steps? For example, I would like to use 3-cycle Morlet wavelet for the
> decomposition, and I would like to have this window slide in time steps of
> 10ms.
>
> 5. For eyeblink reduction method, I understand that it's the most common
> practice to use ICA. I also understand that there are particular pluses and
> minuses of various artifact reduction methods. From my lit review, I
> couldn't find any sort of preference for a method over others in studies
> that looked at coherence, and couldn't find any review on this matter
> either. Does the fact that ICA is a powerful tool justify using ICA for
> artifact reduction in preprocessing data for coherence analysis? Is there
> any issue with this method with particular regards to coherence analysis?
>
> 6. When missing reference channels, would people prefer using a single
> reference (i.e. Cz) or common average reference, in prepping data for later
> coherence analysis?
>
> Thanks in advance.
>
> Best,
>
> Jung
>
> Yonsei University College of Medicine, Korea
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>



-- 
Makoto Miyakoshi
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California San Diego
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20151214/cf3abb7c/attachment.html>