[Eeglablist] ERSP calculation
Makoto Miyakoshi
mmiyakoshi at ucsd.edu
Tue Oct 6 12:48:56 PDT 2015
Dear Michael,
> For a single trial, is it correct that one averages the spectra during
the baseline window (across time) to get the baseline spectrum that is in
the ratio denominator (or used as the baseline subtraction, same thing) for
ERSP calculation?
Interesting quesiton. I actually tested it few years ago. I found that
unless you use very high number of cycles (say above 15-20) you don't get
as sharp spectrum curve as that generated by spectopo(). If you are
analyzing a resting state data or something continuous and don't mind using
a long sliding window, it may worth trying. However, I recommend you
compare the averaged curve over the time with that obtained from spectopo()
to make sure.
Frequency spectrum is a strange thing. If you use Matlab sample code to
compute spectrum, the result looks quite different from that of spectopo(),
but if you use Weltch's method the result is identical (because that's the
method EEGLAB uses).
Makoto
On Tue, Sep 22, 2015 at 3:53 PM, Michael Boyle <mrboyle at live.unc.edu> wrote:
> Hey everyone,
>
> I had a quick question about ERSP calculations. Lets assume my EEG time
> series are simple trials with a baseline period and a period of interest,
> and have been time-frequency transformed through some method of choice and
> the result is a complex spectrogram, from which we extract the amplitude
> spectrogram by simply taking the magnitude. For a single trial, is it
> correct that one averages the spectra during the baseline window (across
> time) to get the baseline spectrum that is in the ratio denominator (or
> used as the baseline subtraction, same thing) for ERSP calculation?
>
> If so, then is it preferred to work with amplitude spectrograms/spectra
> and get dB power by taking 10*log10(amplitude^2/avgBaselineAmplitude^2) or
> to work with power spectrograms/spectra and get dB by taking
> 10*log10(power/avgBaselinePower). These two expressions aren't equivalent
> because mean(baselineAmplitudes)^2 is typically not equal to
> mean(baselineAmplitudes^2) (sometimes very different even), nor is
> mean(baselineAmplitudes)^2/mean(baselineAmplitudes^2) constant, so you
> don't just get a difference of a constant when considering the two results.
>
> Thanks!
> Michael
>
> _______________________________________________
> 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/20151006/3b93d89e/attachment.html>
More information about the eeglablist
mailing list