gammavar is the "standard" variance (ie, the square of the standard error). The residual variance is in rvar.

Mark Hollenbeck wrote:

Hi Doug,

Thanks again for all your help! — Could you explain gammavar.mgh? Would this be considered the stdev of the results found in gamma.mgh?

Thanks, Mark

On Nov 3, 2010, at 11:17 AM, Douglas N Greve wrote:

...

right, that is the correlation, not the correlation coefficient

Mark Hollenbeck wrote:

...

Hi Doug,

Thanks for you help with this. If I am looking for the mean estimated difference (correlation coefficients) between the post-task and pre-task, for certain voxels, would I use gamma.mgh?

Thanks, Mark

On Nov 1, 2010, at 1:50 PM, Douglas N Greve wrote:

...

nope, that's an unpaired test. You can do a paired test in one of two ways.

Method 1: is to create an FSGD file with a class for each subject and one continuous variable. Set this variable to +1 for after task and -1 for before task. Run with DOSS and test the continuous variable. For this test you will need a contrast vector like [0 0 0 ... 0 0 0 1], where the number of zeros equals the number of subjects. In this case, it will best for after>before.

Mthod 2: do your own subtraction after-before, then run mri_glmfit with --osgm. Warning though: make sure that you create a single mask based on all of the data, not on the subracted data.

doug

Mark Hollenbeck wrote:

...

Hi Doug,

I really appreciate your help. I was able to successfully run this analysis on the surface using the --paired-diff method (mris_preproc). To do this in the volume, however, I'm not quite sure if my methods are accurate. I was hoping you could give me some insight. This is what I did:

For volume, I created a 4D image that contained all subject's data (2 images per subject, so 2*N frames). Created a FSGD file with:

... Input Subject1 before_task Input Subject1 after_task Input Subject2 before_task Input Subject2 after_task ... Input SubjectN before_task Input SubjectN after_task

And used glm_fit, with the 4D volume as --y, my FSGD file (with --allowsubjrep), and a simple contrast of [-1 1].

Would this method be essentially equivalent to the --paired-diff surface method in mris_preproc/mri_glmfit? Does it account for intra and inter-subject variance?

Thanks. Again, I appreciate your help. Mark

On Oct 28, 2010, at 1:48 PM, Douglas N Greve wrote:

...

Hi Mark, yes you can. You'll need to run mri_glmfit with an appropriate FSGD file. Do a search for fsgd on the wiki. Let me know if you get stuck.

doug

Mark Hollenbeck wrote:

> Hi Doug, > > I have two groups of resting state data for a subject population: before a task, and after a task. I would like to do a voxel-wise significance test between the two groups using fsfast, or similar, to find the voxels where there is a statistically significant increase/decrease. Can you help? Thanks! > Mark > > >

>

Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html

-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html

-- Douglas N. Greve, Ph.D. MGH-NMR Center greve@nmr.mgh.harvard.edu Phone Number: 617-724-2358 Fax: 617-726-7422

Bugs: surfer.nmr.mgh.harvard.edu/fswiki/BugReporting FileDrop: www.nmr.mgh.harvard.edu/facility/filedrop/index.html