ok, now it's much more clear. Thanks a lot!


On 9 June 2015 at 16:33, H. Nebl <[log in to unmask]> wrote:
It's a general finding, see attachments. If you want to replicate the finding with your own data, look at a very small cluster with just a few voxels and no other sig. voxels attached. Define different VOIs with different radii. As long as you don't cut off any sig. voxels the uncorrected cluster p value remains the same, but the corrected cluster p values are adjusted.

> I should only look a the FWE corrected value, am I right?
The idea of SVC is to adjust corrected statistics according to the (smaller) volume of interest. Which means:
- If you apply a SVC on cluster level you MUST report corrected cluster p values.
- If you apply a SVC on voxel level you MUST look at corrected voxel p values.
Otherwise it's not a SVC in the sense it was invented.