Dear all,
I want to compare the two approaches:
Approach 1: threshold and clusterize at the individual level and then do group level analysis
Approach 2: threshold and clusterize at the group level directly
In general, is one of these better than the other?
About clusterization at the group level - should we average the errts files across subjects and run 3dFWHMx on the resulting avg_errts to get the group-level blur estimate and feed that to 3dClustSim to get the group-level cluster threshold? Or should we just average the individual blur estimates and feed that average to 3dFWHMx? The former sounds like the correct approach for the following reason: When we average the signal across several subjects, we average out the noise and hence the chances of having a noise-cluster at the group level must be lower that the chances of having a noise-cluster at individual level. Hence, we should average the errts files and not just the individual blurs.
Is this understanding correct?
This will be great help!
Best,
Gaurav