3dMVM stats output question

AFNI Message Board
1

Dear AFNI experts,
I wanted to get a bit of information regarding the output of 3dMVM with the following command:

3dMVM -prefix stats.nii -jobs 16 -mask mask.nii \
-bsVars Group+Gender+Age \
-qVars Age \
-num_glt 6 \
-gltLabel 1 CONvsNULL -gltCode 1 'Group : 1*CON ' \
-gltLabel 2 Grp1vsNULL -gltCode 2 'Group : 1*Grp1 ' \
-gltLabel 3 Grp2vsNULL -gltCode 3 'Group : 1*Grp2 ' \
-gltLabel 4 CONvsGrp1 -gltCode 4 'Group : 1*CON -1*Grp1' \
-gltLabel 5 CONvsGrp2 -gltCode 5 'Group : 1*CON -1*Grp2' \
-gltLabel 6 Grp1vsGrp2 -gltCode 6 'Group : 1*Grp1 -1*Grp2' \
-dataTable \

The outputs in stats.nii are Intercept F, Group F, Gender F, Age F, CONvsNULL, CONvsNULL t, Grp1vsNULL, Grp1vsNULL t, Grp2vsNULL, Grp2vsNULL t, CONvsGrp1, CONvsGrp1 t, CONvsGrp2, CONvsGrp2 t, Grp1vsGrp2, Grp1vsGrp2 t

For CONvsGrp1, CONvsGrp2, and Grp1vsGrp2 are the outputs beta values or just the group mean difference?

Thanks
Ajay

2

Ajay,

For CONvsGrp1, CONvsGrp2, and Grp1vsGrp2 are the outputs beta values or just the group mean difference?

They represent the estimated group difference in the same units as the input files.

Gang Chen

3

Hi Gang,
Thanks for the quick reply. If the inputs are resting state correlations (z-score due to fisher linear transform) for a particular seed region, would the magnitude be equivalent to effect size due to the inputs being standardized? If not, is there a way to get the effect size from this data?

Thanks,
Ajay

4

Ajay,

Those sub-bricks in the output represent estimates with the same interpretation as your input. In other words, they are Fisher-transformed Z-values. The term "effect size" can be interpreted differently by different individuals, so provide clarification if you are seeking something other than Z-values.

Gang Chen