[Eeglablist] Using PACT for cross location PAC value computation

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Wed Nov 27 18:34:35 PST 2019


Dear Kolluri,

> I have to compute CFC pac values between electrodes (under the 1020
electrode system), between the various frequency bands for every healthy
control. That is, compute CFC values in the cross location scenario rather
than in the same (sensor) location.

Unfortunately this is not supported in the current PACT.
The product will be ch x ch x phaseFreqBin x ampFreqBin though. Do you have
any idea how to systematically evaluate this? I guess you want to use some
kind of reasonable dimension reduction which I am interested in learning...
I once tried tensor decomposition (PARAFAC) for preliminary test, which
seemed to work ok if the data dimension was not too high.

> 2. A typical timeseries for a control is a 19*(2-5 minutes) worth of rest
state eeg recording. I am wondering, what values for surrogate bins, phase
bins, hfo amplitude percentile will be appropriate ?

Surrogate number could be anything above 2000.
The number of phase bins is a critical parameter because this is directly
used as degrees of freedom in the following inferential statistics. If you
use > 30 then everything will be significant, which seems a bit silly...
sorry this may be a design flaw. Use something between 20 and 30 to make
this statistics meaningful.
Amp percentile is a unique feature of this solution, and it could be just
any value. If you can calculate HFO distribution, you may determine the
value based on it. Otherwise, you can choose traditional p-value threshold
such as 1%, 2.5%, 5%, etc.

If you want to further discuss the possibility of developing cross-channel
PACT, please let me know. We may be able to do it as a collaboration (but
it requires a budget for me though.)

Makoto

On Thu, Nov 21, 2019 at 12:24 AM Kolluri Kanti <kamak.shob at gmail.com> wrote:

> Hi
> I have raw EEG sensor data of many healthy controls from a wide
> demographic.
>
> To compute CFC, I used PACT for this data between various frequency bands
> (using the pop_pac_main gui). I selected sampling pool as 'each channel'
> and kept default values for the rest.
>
> I have to compute CFC pac values between electrodes (under the 1020
> electrode system), between the various frequency bands for every healthy
> control. That is, compute CFC values in the cross location scenario rather
> than in the same (sensor) location.
>
> 1. How to do this with PACT? (cross location cfc)
>
> 2. A typical timeseries for a control is a 19*(2-5 minutes) worth of rest
> state eeg recording. I am wondering, what values for surrogate bins, phase
> bins, hfo amplitude percentile will be appropriate ?
>
> Note that this is rest eeg with NO events.
>
> Thank you
> _______________________________________________
> 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
>


More information about the eeglablist mailing list