You have a single cluster of 148,000 voxels. You can be very confident that this cluster is unusually large under the null hypothesis (cluster P_FWE is <0.001). However, as it probably includes the entire brain, it doesn't offer much information about where the brain you have an effect.
-Tom
On Tue, Jan 31, 2012 at 9:01 AM, SUBSCRIBE SPM Anonymous
<[log in to unmask]> wrote:
Hi,
I am sorry I have others questions about how to use topological FDR!
For seeing my results: I use "none>p=0.001" for the threshold and then an extent of 100 voxels. My results appear! Then I click on "whole brain": a table appear with different informations:
set-level cluster-level peak-level
p c pFWE-corr qFDR-corr KE puncorr pFWE-corr qFDR-corr T Z puncorr
0.000 0.000 148000 0.000 0.064 0.086 4.5 4.42 0.000
At the bottom:
I can read: FWEp: 4.561, FDRp: inf, FWEc: 148000, FDRc: 148000
So I would like to know exactly how I can read my results!
1/ For example I use FDRc and I compare what is the qFDR-corr in the column "cluster-level", isn't it? Here my FDRC is 148000, so my KE is 148000. My results is corrected for topological FDR at the cluster, isn't it? and It is accepted at this threshold isn't it?
2/ I use "None "with 100 voxels. Is it acceptable to use:
- None p=0.005?
-None p=0.01?
-None p=0.05?
Thanks,
All the best,
Charlotte
--
__________________________________________________________
Thomas Nichols, PhD
Principal Research Fellow, Head of Neuroimaging Statistics
Department of Statistics & Warwick Manufacturing Group
University of Warwick, Coventry CV4 7AL, United Kingdom