3dMVM gltcodes question

AFNI Message Board
1

Hi Gang:

I am writing a 3dMVM script for a scan where a participant saw 6 stimuli. Each stimulus was different but there were 3 categories of stimuli (i.e. CSp, CSm, CSt) and 2 exemplars of each. Thus the stimuli were CSm1, CSm2, CSp1, CSp2, CSt1, CSt2.

My Question: Some of the contrasts I am interested are, for example, ALL of the CSms vs. ALL of the CSps (similar to the emotional vs. neutral pictures in the 3dMVM example 1). I have attached my code and attempted to code this in -gltLabel 10-12. Have I done this correctly? Since the codes do not need to add to 0 I just want to make sure I have not messed this up in any way.

Thanks!
Lauren


3dMVM -prefix /Volumes/SEQUEL_DISS/PRO_AFNI/OUTPUT/derivatives/STATS/whole_brain/MVM_TvsC.nii.gz -jobs 4 \
-bsVars "sex+stait+stais" \
-wsVars "cs" \
-qVars "stait,stais" \
-qVarCenters '38.47058824,41.70588235' \
-SS_type 3 \
-num_glt 12 \
-SC \
-gltLabel 1 csm1-csm2 -gltCode 1 'cs : 1*csm1 -1*csm2' \
-gltLabel 2 csp1-csp2 -gltCode 2 'cs : 1*csp1 -1*csp2' \
-gltLabel 3 cst1-cst2 -gltCode 3 'cs : 1*cst1 -1*cst2' \
-gltLabel 4 csm1-csp1 -gltCode 4 'cs : 1*csm1 -1*csp1' \
-gltLabel 5 csm2-csp2 -gltCode 5 'cs : 1*csm2 -1*csp2' \
-gltLabel 6 csm1-cst1 -gltCode 6 'cs : 1*csm1 -1*cst1' \
-gltLabel 7 csm2-cst2 -gltCode 7 'cs : 1*csm2 -1*cst2' \
-gltLabel 8 csp1-cst1 -gltCode 8 'cs : 1*csp1 -1*cst1' \
-gltLabel 9 csp2-cst2 -gltCode 9 'cs : 1*csp2 -1*cst2' \
-gltLabel 10 csm_all-csp_all -gltCode 10 'cs : 1*csm1 +1*csm2 -1*csp1 -1*csp2' \
-gltLabel 11 csm_all-cst_all -gltCode 11 'cs : 1*csm1 +1*csm2 -1*cst1 -1*cst2' \
-gltLabel 12 csp_all-cst_all -gltCode 12 'cs : 1*csp1 +1*csp2 -1*cst1 -1*cst2' \
-dataTable @table.txt
2

I have attached my code and attempted to code this in -gltLabel 10-12. Have I done this correctly? Since the
codes do not need to add to 0 I just want to make sure I have not messed this up in any way.

Those three contrast formulations should be fine as long as they make scientific sense. And the contrast weights do seem to add up to 0 for all of the three.

3

Hello,

I have a related question. I am trying to better understand the differences between a few weight options.

I have a similar contrast in which I originally specified

1happy 1sad -2*neutral (emotion - neutral contrast is the goal)

Would it be equally correct to do

1happy 1sad -1*neutral

if the weights do not need to add up to zero?

I tried it out to see if the results were the same and they were quite different, so I wanted to understand what is happening that is different between these two.
If the desired contrast is emotion> neutral, which model weighting strategy would be more accurate to ask that question?

Thank you!

4

The two inferences are different. The first one

1happy 1sad -2*neutral

compares the sum of ‘happy’ and ‘sad’ with the double amount of 'neutral. In contrast, the second one

1happy 1sad -1*neutral

compares the sum of ‘happy’ and ‘sad’ with 'neutral.

5

Hi Gang,

Thank you so much for your response.
I have a clarification question. If my goal is to compare the average of happy and sad conditions to the neutral condition, is it correct to specify 1happy 1sad -2*neutral?

Specifically, my goal is to have emotion (average of happy and sad) to be weighted the same as neutral.

Thanks again for your help!

6

If my goal is to compare the average of happy and sad conditions to the neutral condition, is it correct to specify 1happy 1sad -2*neutral?

The t-statistic would be fine, but the effect magnitude would be incorrect. So do this:

0.5happy 0.5sad -1*neutral