Hi Gabriel,
for mri_glmfit, W = diag(sqrt(1./yffxvar)), ie, it is a diagonal matrix where each element is the square root of the inverse of the variance from the lower level. This changes the equation from a simple ordinary least squares into a weighted least squares (though not quite a true mixed-effects analysis).
There are not explicitly nuisance regressors in the B matrix. The B is the matrix of regression coefficients ("betas") based on the solution to the forward model y=X*B (B=inv(X'*X)*X'*y). So the frames in B depend on the design matrix which depend on how you've set up your FSGD (or if you've entered the X matrix explicitly).
I don't have another manuscript, but there is a lot of detail in the tutorial slides ftp://surfer.nmr.mgh.harvard.edu/pub/docs/freesurfer.groupanalysis.ppt
doug
Gabriel A. Tobon wrote:
Hello Freesurfers
I would like to run a specialized group GLM analysis on the volume using estimates from the individual subject analysis in Freesurfer as summary statistics. I would like to learn a little more about how this estimation is done before doing so. I have looked at the documentation for "mri_glmfit" and the following tutorial: http://surfer.nmr.mgh.harvard.edu/fswiki/FsFastTutorial I also have your 1999/2003 manuscripts on "Signal Processing and Statistical Analysis for Event-Related fMRI" and "fMRI Analysis 101 - Univariate Analysis," but I am not sure if the method has been updated since then.
- I noticed there is a weight matrix (W) described for "mri_glmfit"
command. How is this derived and used? I can't find more information about it in the documentation/mailing list.
- How are the nuisance regressors organized in the B matrix? I assume
the conditions are first? What are the following regressors?
- Do you have a similar manuscript that describes just how group
analysis is implemented in FsFast?
Thank you for your time and help! Gabriel
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer