[Eeglablist] Seeking power and understanding (newtimef)

Makoto Miyakoshi mmiyakoshi at ucsd.edu
Thu Aug 18 10:23:29 PDT 2016


Thanks you Johanna!

John, there is related discussion going on on the list, which is about the
similar case of using wavelet.

Makoto


On Tue, Aug 9, 2016 at 1:32 PM, Johanna Wagner <joa.wagn at gmail.com> wrote:

> Dear John, dear Makoto,
>
> the FFT output from newtimef() is normalized by window length. timefreq()
> is a function within newtimef() - newtimef() is scaling the timefreq()
> output.
> You can get all timefreq() output values via newtimef(), so I would
> recommend using the latter.
>
> John, newtimef() is dividing by the baseline (by default that is the
> period before 0 in an epoch - if not specified otherwise) - as Makoto wrote.
> Thus the power values are normalized relative to baseline and no scaling
> is needed...
>
> You do not need to use rmbase() as it removes the baseline in the
> timedomain - newtimef() is removing average power at each frequency instead.
>
> Best, johanna
>
>
>
>
> 2016-08-09 11:25 GMT-07:00 Makoto Miyakoshi <mmiyakoshi at ucsd.edu>:
>
>> Dear John,
>>
>> > I think it is percent of baseline?
>>
>> No. If I remember correctly, it is dB change relative to the baseline
>> value. It's usually negative latency, but if data are continuous (which may
>> be your case), it should be whole-epoch value.
>>
>> > What I would really like to have is just µV^2 so I can eventually
>> compare, say, C3 between conditions and see the difference. Is that
>> possible?
>>
>> We just confirmed today that output from timefreq() is NOT normalized by
>> length of data. Which means that if your wavelet window size is longer, it
>> returns larger values. I'm not sure if this is corrected in newtimef()
>> which calls timefreq() but I guess it is so anyways. We started working on
>> this issue, so you'd better wait for our update. Otherwise, if you don't
>> mind coding yourself, you can normalize the 'tf' output from timefreq() by
>> the length of wavelet which you can obtain by using dftfilt3().
>>
>> Makoto
>>
>>
>>
>> On Sat, Aug 6, 2016 at 10:56 AM, John Johnson <john at johnjohnson.info>
>> wrote:
>>
>>> I've combed the interwebs and still haven't found a good answer.
>>>
>>> My data has a -5s rest period followed by a 20s active period (task
>>> related experiment, not event related). My pipeline includes rmbase(),
>>> among other things.
>>> Now, when I run newtimef, I can create nice plots from the ersp obtained
>>> using:
>>>
>>>     [ersp, itc, powbase, times, freqs, erspboot, itcboot, tfdata] = ...
>>>
>>>         newtimef( EEG.data(channel,:,:),...
>>>
>>>             EEG.pnts, [EEG.xmin EEG.xmax]*1000,...
>>>
>>>             EEG.srate,...
>>>
>>>             cycles,...
>>>
>>>             'plotersp', 'off',...
>>>
>>>             'plotitc', 'off',...
>>>
>>>             'freqs', [2 25],...
>>>
>>>             'scale', 'abs');
>>>
>>>
>>> I just do a mean(ersp, 2) to average across time, then mean() across
>>> subjects. The x-axis of my plot is frequency, but I don't know what the
>>> units for the y-axis would be. I think it is percent of baseline? Now if
>>> the values I'm seeing are percent of baseline, that means they're measured
>>> relative to the baseline, which is redundant because I've already removed
>>> the baseline.
>>>
>>> If I add the 'baseline',[NaN] argument to newtimef, my plots then look
>>> like ski slopes (high at low frequencies, low at higher frequencies).
>>>
>>> What I would really like to have is just µV^2 so I can eventually
>>> compare, say, C3 between conditions and see the difference. Is that
>>> possible?
>>>
>>>
>>> I've seen this question several times on the 'net, but haven't found a
>>> good answer.
>>>
>>>
>>> Thanks,
>>>
>>> John
>>>
>>>
>>>
>>>
>>> --
>>> Sent from a MacBook Pro
>>>
>>> _______________________________________________
>>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.uc
>>> sd.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
>>
>> _______________________________________________
>> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
>> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.uc
>> sd.edu
>> For digest mode, send an email with the subject "set digest mime" to
>> eeglablist-request at sccn.ucsd.edu
>>
>
>
>
> --
>
> Johanna Wagner, PhD
> Department of Psychology,
> University of California San Diego
> http://aronlab.org/people/
> http://scholar.google.at/citations?user=vSJYGtcAAAAJ&hl=en
>
> <http://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/20160818/dba4ac2a/attachment.html>


More information about the eeglablist mailing list