Comparing to zero

Tom Holroyd tomh at KURAGE.NIMH.NIH.GOV
Mon Jun 27 16:44:38 CEST 2005

Dr. Jose Luis Patino Vilchis wrote:
> We are also working with coherences. I would be interested to know more about
> the algorithm embedded in the function and to know how you use it to find
> significant coherence. Could it bepossible to have more information on this?

At the moment the best way I've found is as follows:

Compute 'powandcsd' with freqanalysis().
Use the coh_refdip method of sourceanalysis().
Save the avg.coh array of the result per subject.
Using AFNI, I then take the sqrt() and z-score each volume.
The result is reasonably normally distributed.
Warp to Talairach space.
The final step is to use 3dANOVA2 with a mixed effects model,
with factor A as the conditions in the experiment, and factor
B as subjects.  The most important step here is that the results
are computed as contrasts between different conditions.  Otherwise
there is a huge (quarter brain) coherence blob around the reference
dipole.  It usually subtracts out pretty well in a contrast.

Dr. Tom Holroyd
"A man of genius makes no mistakes. His errors are volitional and
are the portals of discovery." -- James Joyce

Note: if tomh at kurage doesn't work try holroydt at

More information about the fieldtrip mailing list