Hi SPM
Just checking on my understanding of the variation between the degrees of
freedom of a model and the estimated smoothness of the residual random
fields and resulting resel calculations. From my understanding and some
empirical measurements (below) it appears that as the degrees of freedom
falls, the residual images get rougher, FWHM decreases, voxel per resel
decreases and the number of resels increases.
Is this correct and to be expected?
Example (The following analyses are 2nd level fMRI analyses; all use the
same images, and end up with the exact same mask image. The model varies by
including different main and interaction effects, which affects only the
degrees of freedom)
model totVox (in mask) DOF FWHM VpR Resels
1 55227 38 [3.78 3.76 3.59] 51.03 966.9
2 55227 40 [3.80 3.79 3.61] 52.05 948.0
3 55227 42 [3.95 3.95 3.79] 59.25 832.7
thanks,
darren
---------
Darren Gitelman, MD
Department of Neurology
Northwestern University
voice: (312) 908-8614
fax: (312) 908-5073
----------
|