[Eeglablist] Time-frequency analysis (subtraction first or analysis first)

[Thomas Ferree]

Analysis of EEG time series falls under the heading of stochasticsignal
processing.  There is an understanding that, even if the system
response properties, i.e., frequency-domain transfer function, are the
same across trials, the exact data recorded will be different across trials.

Statistical estimation of the power spectrum always involves some
form of averaging.  See Numerical Recipes in C, or Spectral Analysis
for Physical Applications by Percival and Walden.  When we are fortunate
enough to have multiple data samples, e.g., trials in cognitive tasks,
then we average across trials.  This averaging can take place in each
short moving window, as is done in time-frequency analysis.

So to look for differences in spectral response properties across trials,
the proper procedure is to estimate the power spectrum in each condition
(and each moving window) by averaging across trials, then subtract the
two conditions in each window, as Stan and Julie said.  That is what
newtimef does, although EEGLAB also handles the baseline interval,
which I have not addressed here.

I send this comment mainly to point out that simulations with deterministic
sine waves are not representative of the situation with stochastic signals
like EEG data.

-- 
Thomas Ferree, PhD
Department of Radiology
UT Southwestern Medical Center
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20080415/b483e4f1/attachment.html>