coherence again

[Jan Hirschmann]

Dear fieldtrip experts,

I have yet another question about the statistical testing of coherence.
Again, it relates to the group level, this time to the testing procedure
used in "Imaging the motor system's beta-band synchronisation during
isometric contraction", Schoffeleln et al., NeuroImage, 2008.
There it says: " The null hypothesis that was tested at the voxel level
states that the Z-transformed coherence between the EMG and that voxel,
accumulated across subjects, is not different from 0. Under this null
hypothesis, flipping the sign of the Z-transformed coherence values of a
random subset of the subjects before accumulating leads to an
alternative observation that belongs to the null distribution."

Now, as I understand it, the Z-transformation described here is

z=sqrt(-(df-2)*log(1-abs(C)^2)), i.e. z takes only positive values
(there is no comparing of conditions here)

And accumulation, it later says, means summing up the z-values of all
subjects and dividing by sqrt(n) at each voxel. I suppose n is the
number of subjects.

So "flipping the sign" means multiplying the z-value of some subjects by
-1 before summing up? As all z-values were originally positive, this
will inevitably decrease the sum. So the original state will inevitably
be the extreme end of the distribution.

Sorry, I must have some serious error in my thinking here (hope it's not
too obvious). Can you help me?

Many thanks,

Jan

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.