85,272,240,317,229,412,000,000,000 Resels!? That's a big search volume! Lots of Resesl == Tiny FWHM == Gigantic roughness. You probably have a overly generous mask that includes non-brain voxels which have nearly no smoothness.
Check the RPV, FWHM images for problems. See message from yesterday,
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0810&L=SPM&P=R12991 for more.
-Tom
On Tue, Oct 7, 2008 at 4:36 PM, Veronica Garcia Vazquez
<[log in to unmask]> wrote:
Dear List,
We have problems with p-values (corrected ones in set-level and in cluster-level) in a two sample t-test. Theirs values are NaN and we don´t know what they mean żnot a number???. I would appreciate any kind of help.
We are running SPM5 (version 1782) in Matlab 7.0.
Our design and parameters are the following ones:
1))) In SPM5, we´ve chosen a Two Sample t-test.
Group 1: Controls (11 images).
Group 2: Patients (8 images).
Independent Yes and Variance UnEqual.
The other options are the default ones.
2))) No problems with estimating parameters.
3))) Results:
Contrast Vector: 1 -1
No mask with other contrasts
p-value adjustment to control: NONE
threshold: 0.005
Extent Threshold: 10 (voxels)
In the email is attached the p-values (whole brain) window with the NaN values.
We've realised that expected voxels per cluster, expected number of clusters, FWHM and resel size are zero.
Thank you so much.
Regards,
Elena
Veronica
--
____________________________________________
Thomas Nichols, PhD
Director, Modelling & Genetics
GlaxoSmithKline Clinical Imaging Centre
Senior Research Fellow
Oxford University FMRIB Centre
