At 03:07 PM 4/28/2005 -0400, Huang, Chaorui wrote:
>I got some weird findings. It would be very helpful if someone could give
>some hints. This is a very simple two group t-test using SPECT data. But
>the corrected cluster level turns to be significant when it contains only
>one voxel. Is it because there are some problems with my SPECT images or I
>made a wrong study design? Can the results be used? I would be very
>pleased if anyone could help.
One reason I could think of is that your threshold is too high. Are you
using the "FWE p<0.05" threshold? If so, you might want to lower the
threshold (say, p<0.001, or p<0.0001 uncorrected). I think that should
produce more "reasonable" cluster-level p-values. If you want to correct
for multiple comparisons at the voxel-level, then I wouldn't worry about
cluster-level p-values though.
FYI, a more detailed explanation about a high threshold and its impact on a
cluster size test is found in the SPM mailing list archive:
http://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=ind04&L=spm&P=R460921&I=-1
-Satoru
Satoru Hayasaka ==============================================
Post-Doctoral Fellow, MR Unit, UCSF / VA Medical Center
Email: shayasak_at_itsa_dot_ucsf_dot_edu Phone:(415) 221-4810 x4237
Homepage: http://www.sph.umich.edu/~hayasaka
==============================================================
|