[Eeglablist] Extract absolute power from newtimef

[schmiedt at uni-bremen.de]

Dear EEGlab community,

In order to do statistics on the time frequency information of EEG  
data, I am interested in the absolute power of the signal in units of  
squared microvolt.

As far as I understand the documentation and the information on the  
mailing list I have to do the following: use the 'no baseline'  
parameter for the newtimf command and convert the values from decibel  
to absolute values.

To test this I used the attached code on Matlab 7.5 and the current  
EEGLAB toolbox. It replaces the first channel of the current data set  
with a sinusoid testing signal (frequency=20Hz, amplitude=5µV ). Then  
the above is done.

Some resulting power values of the signal at 20Hz are:

  a P
  1 1.0703e+04
  2 4.2811e+04
  3 9.6325e+04
  5 26.757e+04

The leading numbers seem okay to me, but I don't understand the order  
of the power. What is the unit of the resulting power values? And why  
isn't it (µV)²?

I would be very glad if anyone could explain this.

Best regards,
Joscha Schmiedt



% create testing signal with f=20Hz and a=5
t = linspace(EEG.xmin, EEG.xmax, EEG.pnts);
testsignal = (ones(EEG.trials,1)*5*sin(2*pi*20*t))';
EEG.data(1,:,:) = testsignal;

% compute and plot ersp
figure
[ersp,itc,powbase,times,freqs,erspboot,itcboot] = newtimef( ...
    EEG.data(1,:,:), ...
    EEG.pnts, ...
    [EEG.xmin EEG.xmax]*1000, ...
    EEG.srate, [3 0.5], ...
    'baseline', NaN);

% get maximum power
[erspmax, i] = max(ersp);

% results
Power = 10^(erspmax(1)/10)
Frequency = freqs(i(1))