[Eeglablist] Power spectrum too high?

[Thomas Ferree]

Dear Michael,

There are a variety of normalizations for power spectra as matters of
definition.  In addition, tapering, zero-padding, and using one-sided versus
two-sided conventions can change apparent values further.   Numerical
Recipes in C has a nice summary of these issues.  You can access freely by
following this link http://apps.nrbook.com/c/index.html, then go to Section
13.4, page 550.

Consequently, comparing numerical values of power spectra across studies is
difficult and normally avoided.  Instead, most researchers restrict their
efforts to test for group differences, task differences, etc, within
whatever normalization conventions they are using within their lab or study.

In the past I have scrutinized the calculation of power spectra in EEGLAB
and everything is done correctly, so you need not worry that something is
wrong based upon the numerical values you are seeing.

Best, Tom.

On Sat, Feb 5, 2011 at 11:26 AM, Michael Gandal
<mgandal at mail.med.upenn.edu>wrote:

> Hi,
>
> I use the FFT power spectrum function for a single channel and it outputs a
> nice spectrum in dB that looks qualitatively very much like it should.
> However, when I compare with the values that I calculate by hand with
> matlab or Spike2 it seems that the EEGlab power spectral values are always
> too high.
>
> Does anyone know why this is and how to correct for it?
>
> Thanks
> Mike
>
> _______________________________________________
> 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
>



-- 
Thomas Ferree, PhD
Cell: (415) 577-1285
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110211/c16fe8a8/attachment.html>