I'm still working on the docs :), but in answer to your questions:
1) X is the design matrix. 2) Yes, you must look at the fwhm in the output log file and use that value.
There's more info if you run mri_glmfit with --help.
here are some refs:
Statistical Science 2004, Vol. 19, No. 4, 676-685 DOI 10.1214/088342304000000396 © Institute of Mathematical Statistics, 2004 Permutation Methods: A Basis for Exact Inference Michael D. Ernst
Human Brain Mapping 15:1-25(2001) Nonparametric Permutation Tests For Functional Neuroimaging: A Primer with Examples Thomas E. Nichols1 and Andrew P. Holmes
doug
zvalim@gmail.com wrote:
Dear FreeSurfer mailing list,
I have looked at the online documentation for the new version, and was wondering where one could find more details on the permutation algorithm implemented in the current version. Some issues I would like to understand better is how the smoothing in the dataset is captured in the algorithm, and what is being permutated at the group level. If there is any reference to a published paper implementing this exact algorithm it would be very useful as well.
Specifically, what I found online was: The first parameter the nulltype, which is the method of generating the null data to be tested. Useable options are: (1) perm - perumation, randomly permute rows of X (cf FSL randomise) (2) mc-full - replace input with white gaussian noise (3) mc-z - do not actually do analysis, just assume the output is z-distributed (cf ANFI AlphaSim)
I have two questions of this... In (1 perm), what does "X" stand for -- a conditionXsubject table for each voxel? In (2 mc-full), when WN gaussian is inputted, is it smoothed with the characteristics of the original dataset?
Sincerely, Uri
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer