[Eeglablist] Permutation test for ERSP?

[Tero Hakala]

>>  I have understood that a fairly straightforward way to do this might be a
>>  permutation test (or bootstrap random draw?)  where one randomly assigns
>>  subjects to 2 groups, calculates the group averaged ERSP and their
>>  differences, repeat this for many times and end up with a distribution in
>>  each TFR-point. Finally use these distributions to look for statistically
>>  significant points between the ERSP images of my two groups.
>
> If you look at the help message of the newtimef() function, that is exactly 
> what it does. You can only compare two conditions though.

Thank you, it seems to work nicely. One question though: in condstat.m 
the significance thresholds are set to the mean of i (naccu*alpha) 
lowest or highest values i.e.,

accarraytmp(:,:,2) = mean(accarray(:,:,1:i),3);
accarraytmp(:,:,1) = mean(accarray(:,:,naccu-i+1:naccu),3);

why use mean and not simply the i:th lowest/highest value?

  -t