[Eeglablist] About ERSP averaging and normalisation

[Marco Buiatti]

Dear eeglab people,

I have some questions about how to normalise and average ERSPs across 
electrodes and subjects. I saw that eeglab performs the following steps:
a) for every electrode, subject and frequency, normalisation is 
performed by dividing the ERSP by the mean ERSP along the baseline. 
Then, 10log10 is taken.
b) Bootstrap analysis sets a threshold for statistical significance.
c) Average/RMS is taken over all subjects and electrodes (optionally 
with another threshold on the minumum number of significant 
electrodes/subjects).

Here the questions:
1) Why not normalising by subtracting the mean and dividing by the std 
of the baseline? Intuitively, I guess this has something to do with the 
fact that ERSP>0, and your normalisation preserves this, but I cannot 
find a deeper explanation.
2) The non-linearity of the log causes averages over electrodes/subjects 
to be biased: don't you think this can be misleading?
3) What is the advantage to use RMS instead of averaging? On one hand 
RMS enhances differences (w.r.t. baseline), but on the other it could 
enhance noise. Again, is there any deeper explanation?
4) This is more a prayer than a question, since the question has been 
asked several times: a tool to quantify the statistical difference 
between two conditions would be highly appreciated!

Little thing to correct: when using tftopo, 'mode','ave', the message 
'RMS averaging' is printed, while it seems to me that normal averaging 
is performed.

Thank you,

Marco

-- 
Marco Buiatti

Neurophysique et Physiologie du Systeme Moteur - CNRS UMR 8119
UFR biomedicale Les Saints-Peres
45 rue des Saints-Peres 75270 Paris cedex 06
Tel: +33(0)142862146
Fax: +33(0)149279062
http://www.neurophys.biomedicale.univ-paris5.fr/~buiatti/