3dMVM centering

Hi all,

Question about centering in 3dMVM. We have a prior concussion variable (def1/def2 below) of interest that we’re trying to run without mean centering. However, after running one 3dMVM with qVarCenters and one without, the output/results appear to be the same, i.e., it appears the default mean centering occurs regardless of whether qVarCenters is used. The actual mean is around 1, but we want to account for values of 0. Are we missing something in our code to account for this?

(Version AFNI_19.3.18)


3dMVM -prefix ${resDir}/cbfMVM_def2_Model2_noMC \
    -jobs 24 \
    -resid def2_Model2_noMC_resid \
    -bsVars 'def2*sex+total_years_exposure+ageAtScan+waistHipRatio' \
    -qVars 'def1,def2,total_years_exposure,ageAtScan,waistHipRatio' \
    -qVarCenters '0,0,11.27,21.50,0.881' \
    -mask ${homeDir}/data/cbf/analysis/MASK/CBF_inter_GM+tlrc \
    -num_glt 2 \
    -gltLabel 1 men_con -gltCode 1 'sex: 1*0 def2 : ' \
    -gltLabel 2 women_con -gltCode 2 'sex: 1*1 def2 : ' \
    -dataTable \
    Subj def1 def2 ageAtScan total_years_exposure sex waistHipRatio InputFile \


Output example (based on code from above):
96 centered values for numeric variable def1 : 0.65625 0.65625 3.65625 1.65625 2.65625…

Output example (with default mean centering):
96 centered values for numeric variable def1 : 0.65625 0.65625 3.65625 1.65625 2.65625…

Perhaps it’s a small enough difference to not matter? Not sure. Thoughts?

Thanks,
Lezlie

Lezlie,

With your model defined as:

-bsVars ‘def2*sex+total_years_exposure+ageAtScan+waistHipRatio’ \

def1 is not even part of the model. Or, I’m missing something?

Hi Gang,

That’s correct. We’re running a few different models, but swapping out def1 and def2 to compare two different definitions. They’re both in the data table so we include both in the qVars but only one definition at a time goes in the actual model.

-Lezlie