On 8/8/18 2:17 PM, Funk, Quentin wrote:
External Email - Use Caution
Freesurfer experts,
I'm trying to figure out what values I get from mri_glmfit as in the tutorial. https://surfer.nmr.mgh.harvard.edu/fswiki/FsTutorial/GroupAnalysisIn particular, I'm interested in how I can compare these values to what I get from QDEC.
QDEC runs mri_glmfit, so they will be the same assuming you give it the same input, design, and contrast.
In that tutorial, the software does a linear regression for each gender and then computes a F_ratio for the two two fits. (How does this relate to the average in the contrast listed?)
If the average were 0, then you would have no result whereas the F test will test for any combination of the two
The F_ratio then gives a way to get P-values (under certain assumptions)--so then from the F_ratio, pvals are computed--this leads to sig.mgh.
But what are the other files? The descriptions are a bit too vague for me to understand, unfortunately. The ones I'm interested in primarily are:
pcc.mgh -- partial (pearson) correlation coefficient (surface overlay)
Do you know what a pearson correlation is? The pearson can only be computed between two sets of numbers. When you have multiple regression, you use a partial pearson.
z.mgh -- z-stat that corresponds to the significance (surface overlay)
This is the p-value converted into a z.
The z-stat should be some sort of standard score--so a standard score of significance (so p-value?) compared to what distribution?
Second, what if I were interested in a different categorical variable--say I had a group of Alzheimer's patients and a group of Healthy patients. In QDEC, I click "diagnosis" as the variable to do the analysis on. What contrast does this correspond to? What type of linear fit is being compared--the slope of the regression with respect to age? If I leave out nuisance variables (like age), I should conceivably get an F-test between the two constant models, which should be more or less similar to a voxelwise t-test?
The contrast will be [-1 +1]. The sig.mgh will be the sig of the t-test (ie, a signed test).
Any help here is appreciated!
Quentin Funk, PhD Houston Methodist Research Institute 713-363-9003 tel:713-363-9003
Houston Methodist. Leading Medicine.
Houston Methodist is ranked by /U.S. News & World Report/ as the No. 1 hospital in Texas for patient care. Houston Methodist is nationally ranked in 8 specialties and is designated as a Magnet hospital for excellence in nursing. Visit us at houstonmethodist.org <houstonmethodist.org>. Follow us at twitter.com/MethodistHosp <twitter.com/MethodistHosp> and facebook.com/HoustonMethodist.
***CONFIDENTIALITY NOTICE*** This e-mail is the property of Houston Methodist and/or its relevant affiliates and may contain restricted and privileged material for the sole use of the intended recipient(s). Any review, use, distribution or disclosure by others is strictly prohibited. If you are not the intended recipient (or authorized to receive for the recipient), please contact the sender and delete all copies of the message. Thank you.
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer