Hi Wayne, this is the problem with these types of corrections. As you change the parameters, the null distribution for cluster sizes change. At a threshold of .01 you may get a cluster that is smaller than at .05, but it may be more significant because the shape of the null distribution changes. One parameter set is not more "right" than another. Sorry, I wish I had more guidance for you. doug
On 07/15/2013 01:10 PM, Wayne Su wrote:
Confused results from Qdec Dear Freesurfers,
I used Qdec to do two-group comparison on cortical thickness. Demeaned age was used as the covariate. The question I’m looking at is that “Does the thickness-age correlation differ between C1 and C3?”
When FWHM=15mm was chosen, I got confused results: When I used threshold 1.3(0.05) in Monte Carol Null-Z Simulation, I didn’t get any significant cluster. However when I used threshold 2(0.01), I did get one significant cluster. However I expected I at least got same significant area when I used lower threshold.
If I chose FWHM=10mm, I got another significant cluster when the threshold 1.3 was used, and got nothing when the threshold 2 was used. However this cluster is not the same area I got when FWHM=15mm was used.
At this point, I’m totally confused, and don’t know which one should be trusted. Should I used FWHM=10 or 15? Should I used threshold 1.3 or 2.0?
My group one have 49 subjects and group two have 46 subject. All subjects are males.
Hope someone can point the right direction for me. Thank you very much.
Wayne
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer