I am working on a VBM data set trying to transform T values to r coefficients
with the Gaserīs code "cg_spmT2x.m" (118 2009-03-25) after a multiple
regression analysis (two variables included, n=16, df=13). To visualized and
extract voxel-by-voxel r values I use the display toolbox in SPM5. Statistical
T maps are threshold at p<0.001 (cluster threshold p<0.05 FWE corrected,
non-isotropic correction, yes). The problem is that I am getting some kind of
"voodoo" correlation coefficients in the r_images as high as for almost every
voxel r=1 (for example for a T value of 6.67). However, when I directly apply
the formula for r as written in Gaserīs code I get a r value of 0.87.
That is: r= (sign(6.67) / (sqrt(((13)/(6.67*6.67))+1)))
Probably the code is getting some information further into matrix design , but
when I crumble the code it seems to correctly read the df and I donīt get any
error message or unknow value. Of course the sample is small.
I was just looking for age related effect in my sample group to report those
effects. Furthermore, another guess related to age effects in VBM data is that
if I want to look for local gray matter changes associated with age and
independently of individual differences in TIV I should regress out TIV values.
Does this last make sense to you?
Please, any kind explanation to these results will be wellcome.
Thanks in advance.