When you get p-values (sigs) out of FreeSurfer, the p-value is computed assuming an unsigned test. This makes the p-value twice that of a signed test (ie, it is less significant than the signed test). When you go to correct for multiple comparisons and you specify a sign (positive in you case), mri_glmfit-sim recognizes that you are performing a signed test and changes the threshold to be appropriate for a signed test. In your case, your threshold was 2 which means p<.01. The threshold will be changed to p<.02, -log10(.02) = 1.7. So if you go back and check the size of your cluster with the first method and use a threshold of 1.7, you should see that it is the same size as what mri_glmfit gives you.
doug
Irene Altarelli wrote:
Dear Freesurfers,
I am doing a vertex-by-vertex analysis on a label, using mri_glmfit. As a result, I got a very small cluster (62 vertices) at p<0.01. Then I tried running clusterwise correction using the following command:
mri_glmfit-sim \ --glmdir fMRIfaces20.glmdir \ --sim mc-z 5000 2 mc-z.pos \ --sim-sign pos \ --overwrite
I am a bit surprised by the result: one cluster is reported in mc-z.pos.sig.cluster.summary, and it is bigger than the one previously identified.
Could this be correct? if not, what could I be doing wrong?
Please find attached mc-z.pos.sig.cluster.summary and mc-z.pos.j001-lh-try-ctr-dys2.csd.
Thanks very much in advance, Irene
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer