[FieldTrip] Trialwise log of TF transform before condition averaging?

[George Wallis]

Dear fieldtrip experts

When computing condition averages of TF transformed data, I take the log 
of the trialwise data before averaging over the trials for a given 
condition.  As I was taking the log of the trialwise data for trialwise 
analyses, this seemed the most consistent approach. However, I notice it 
isn't implemented in fieldtrip (it requires a tweak of the 
ft_freqanalysis function), and the tf averages tend to look noisier this 
way than if they are averaged before taking the log.

Does anyone have any insight into whether this is a sensible way to 
generate condition averages?  I originally thought this would reduce the 
influence of outlier trials - but since it is equivalent to taking the 
log of the product of all the trials within a condition, perhaps it is 
going to do the opposite?

George