Dear SPMers,
If I were to create a histogram of a (fixed-effects) t-statistic image
(across voxels), I would (naively) expect to see a Gaussian, with perhaps
some extra stuff in the tails representing signal (if any).
When I actually do this, I do get something fairly Gaussian (if I exlude the
0s/small values outside the brain), but it's not mean 0 variance 1. The
mean is often 0.5 (say) and the variance is often about 3 or 4.
Where does this behavior come from? I'm guessing it means the model is
somehow incomplete, but I'm not sure.
(This is fMRI data.)
Best,
Steve Fromm, PhD
NIDCD/NIH
|