Dear Bing,

Cluster inference works best when you have a search volume that is large relative to the smoothness.  For just a 10mm radius SVC region, I wouldn't expect cluster inferences to be of any use at all.  For such small regions I would stick with peak/voxel inferences.

-Tom

On Thu, May 27, 2010 at 7:38 PM, Bing Ye <[log in to unmask]> wrote:
Dear SPMers:
 
I want to use small volume correction (SVC) for the brain activation level. I have encountered an issue, after using the SVC, the puncorrected value at the cluter-level is higher than before (e.g. p=0.06 after using the SVC , whereas before is 0.02. ) , but the p corrected-level is significant after SVC at cluster-level. I do not know the reason, is it supposed to be like that some times?
 
What should I be looking at after SVC, cluster-level p uncorrected or p corrected or voxel-level p fwe-corrected or p fdr-corrected, and at what p level ( p = 0.05)?
 
I entered 10 mm for the radius value
 
Can anyone let me know?
 
Thank you in advance.
 
Bing



--
____________________________________________
Thomas Nichols, PhD
Principal Research Fellow, Head of Neuroimaging Statistics
Department of Statistics & Warwick Manufacturing Group
University of Warwick
Coventry  CV4 7AL
United Kingdom

Email: [log in to unmask]
Phone, Stats: +44 24761 51086, WMG: +44 24761 50752
Fax:  +44 24 7652 4532