Info on doing the paired t is here:
https://surfer.nmr.mgh.harvard.edu/fswiki/PairedAnalysis
Don Hagler wrote:
You could extend the analysis described on that wiki page by calculating tstats and pvalues (in addition to mean and stdev). Then you have a paired t-test, whereas mri_glmfit with qdec would be doing an unpaired t-test. I think preproc has an option to calculate within subject differences for a paired t-test with mri_glmfit as well. As you may know, paired t-tests are more sensitive than unpaired t-tests, and are generally more appropriate when you make within subject comparisons. On the other hand, if you intend your exercise to be a test for the likelihood of false positive differences between groups, the unpaired t-test may be more appropriate.
Subject: Re: [Freesurfer] qdec thickness and repeatability From: nicks@nmr.mgh.harvard.edu To: bell0368@umn.edu Date: Fri, 20 Feb 2009 15:15:10 -0500 CC: freesurfer@nmr.mgh.harvard.edu
The two methods are very different. In the method described on the thickness repeatability page, its computing a mean and stddev of the thickness for your group. This a direct and easily interpretable measure, but doesn't give you a statistical measure (but you could do that on your own with some external package).
In qdec (mri_glmfit), it would be conducting a GLM-based analysis looking for statistical significance, expressed as a p-value at each vertex, in thickness between groups. Here, raw thickness differences are hidden in the analysis.
However, you can of course do both analysis, and compare them, which I would suggest you do. You would hope the areas of change are the same. Note that you can load external surface maps, such as a thickness- diff.mgh map, onto the fsaverage surface in qdec, so that you can view both analysis (or just load the thickness-diff.mgh map in tksurfer).
Its on the list of things to add to qdec to create two-group thickness difference maps, as well as just viewing the mean group thickness, to save some work for this sort of thing.
Nick
On Fri, 2009-02-20 at 12:53 -0600, bell0368@umn.edu wrote:
All,
I am going to run an analysis of cortical thickness
reproducibility; there
are at least two methods that seem obvious. I could run qdec
analysis with
time1 and time2 as different groups or I could run the methods
below. Are
the commands below the same ones used in the qdec GUI? As far as I
can
tell, they will both give me cortical thickness differences on a
vertex by
vertex basis. There are probably some subtle differences in the
way they
are coded, but any major changes? Thanks.
http://surfer.nmr.mgh.harvard.edu/fswiki/ThicknessRepeatibility
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
Windows Live™ Hotmail®:…more than just e-mail. Check it out. http://windowslive.com/explore?ocid=TXT_TAGLM_WL_t2_hm_justgotbetter_explore_022009
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer