SNFR as voxel-wise covariate in 3dMVM

I am using the SFNR output as a voxel-wise covariate in 3dMVM, partly because my study includes multiple scanners, and I want to make sure that cohort differences aren’t being affected by different SNR at different sites.

I can add the sfnr as a nuisance variable and it runs just fine, however if I try to add a contrast including SFNR (by adding, e.g. -gltCode 50 'Group 1*site1 sfnr : ’ ) then I get the error

** Error:
Incorrect level coding in variable sfnr : :

Is it possible with the current 3dMVM program to get a contrast map including the voxel-wise variable?

Could you provide your 3dMVM script (with a few lines in the -dataTable, and with only one GLT for -gltCode 50 'Group 1*site1 sfnr : ’ )?

-prefix output_mvm_2016_10_19_16
-jobs 6
-mask TT_Mask.nii.gz
-bsVars 'Groupsfnr’
-wsVars 'Lex
-vVars ‘sfnr’
-num_glt 1
-gltLabel 1 el1:sfnr -gltCode 1 'Group : 1
el1 sfnr : ’
-dataTable @/path/to/table/data_table.txt

with the data table looking something like this (there are more behavioural variables in there, but they’re not used in this analysis):
Subj Group Mod Lex L1L2 sfnr InputFile
hu_001 el1 prnt wrd L1 /path/to/subj/hu_001_SFNR+tlrc.HEAD /path/to/subj/stats.hu_001_REML+tlrc.HEAD[english_print_wrd#0_Coef]
hu_001 el1 prnt wrd L2 /path/to/subj/hu_001_SFNR+tlrc.HEAD /path/to/subj/stats.hu_001_REML+tlrc.HEAD[hebrew_print_wrd#0_Coef]
hu_001 el1 prnt psd L1 /path/to/subj/hu_001_SFNR+tlrc.HEAD /path/to/subj/stats.hu_001_REML+tlrc.HEAD[english_print_psd#0_Coef]
hu_001 el1 prnt psd L2 /path/to/subj/hu_001_SFNR+tlrc.HEAD /path/to/subj/stats.hu_001_REML+tlrc.HEAD[hebrew_print_psd#0_Coef] \