Dear Alexander,
The cluster-based statistics account for all of these issues.
If you are running one of the cluster-related statistical inferences (from FEAT
or randomise) then anything that survives is "meaningful" in that it is statistically
significant.
All the best,
Mark
On 29 Jul 2011, at 23:47, Alexander Olsen wrote:
> Hi,
> I have now run the cluster command on my voxel based results after they have been registered to a 1mm standard template. The list of clusters are quite long, and some clusters are relatively small as compared to others. My question is whether there is a logical way of setting a cut-off for which (meaningful) clusters to report, i.e. based on image resolution, applied smoothing etc. My first thought is that I should only report clusters that are larger than 216 voxels as that would be the size of one "functional voxel" after smoothing (6x6x6mm)... but I have a suspicion that this may not be correct. Hope someone can guide me on this issue.
>
> Thank you,
>
> Alexander
>
|