Dear David,
> One question I'm still not sure of, however: Is it possible to make variance
> maps in SPM99 and test their equivalence across the brain (one assumes on a
> pixel x pixel basis)? I gather that it should be through one or another
> manipulation of Matlab code (i.e., variance is a step toward a t-test), but
> I don't know much about Matlab.
The variance maps are contained in the ResMS.img that gets written out after
your SPM analysis. The square root of values from the ResMS map are multiplied
by a constant to form the divisor for the SPM (voxelwise variance) T map (where
the con*img is the numerator). You could I guess look at these ResMS values
and get an idea of the variation of variance. I think Keith Worsley looked at
variance in relation to baseline global signal level, and according to distance
from the centre of the slice, and found no striking differences in voxel
standard deviation.
If you need matlab code for this kind of thing, the principles of getting the
data out are in fact all there in the function I posted previously. Please let
me know if you need some exposition and I'll try and provide it.
Best wishes,
Matthew
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|