Although I have some experience with fMRI, I'm very new to SPM so
apologies if this is a silly question.
I want to estimate the smoothness of some images but get weird results
for the estimates. When using spm_smooth to smooth single random
gaussian images with different kernel sizes, subsequent estimates give
much larger values than expected.
Please see this code snippet:
spm_smooth(Y,Ysmooth,[1 1 1]*blur);
V.fname=newfn; % volume
M=V; % mask
Running this script gives values up to 6*10^5 for a smoothness of 12.
Ok, somewhere in the documentation it says it estimates the variance,
so taking the square root seemed like a good idea; this gives indeed a
more or less linear relationship, but the scale is off:
56.0264 51.1458 52.0927 53.7166 56.7196 58.8658 61.0850
61.0351 64.4306 66.0386 68.0434 64.4542
This is of by a factor of 60 or so (or 20, if we consider the voxel
size, which is 3mm isotropic). Any idea what is causing this? What is
the proper way to estimate this smoothness of such individual,