3dMVM questions

AFNI Message Board
1

Dear AFNI experts,

I’m comparing different resting state measures between 3 patient groups and controls. I have to covariate a parameter of the movement of the subjects in the ANOVA I performed between my groups of subjects (4 groups) and in the t-tests comparing these groups (I used 3dANOVA before). The code I used is below. I have the following questions:

  1. is 3dMVM the right choice?
  2. is the syntax in gltcode and glfcode correct for my purpose?
  3. if I look at the results of the MVM on the brick showing only the effect F of “Group” variable I don’t get exactly the same results as the ANOVA without the covariate I had done before using 3dANOVA (they’re similar, clusters are significantly smaller in 3dMVM); are they supposed to be different?
  4. I read the guide about multiple centers for quantitative variables and concluded that it might be better to choose multiple centers (the mean Movement in the groups) to account for the possible interaction between group and movement, even if the interaction between group and movement is not significant tested in an ANOVA. I also tried to use a single center for the Movement variable (I did not put the qvarcenters option) and I got exactly the same results as for multiple centers. Might this be because the groups don’t differ enough in Movement? Should the centers in the qvarcenters options be written in a specific order?
  5. In my case, is it the same to put the quantitative variable Movement inside glfcode and gltcode or not? I got exactly the same results trying both alternatives.

The results I obtained are a bit problematic and I’d like to be confident that I ran the analyses correctly.

Kind regards,
Daniel



3dMVM -prefix MVM_GROUP_MOVEMENT -jobs 8 \
-bsVars "Group*Movement" \
-qVars "Movement" \
-qVarCenters '(0.0450,0.0607,0.0575,0.0492)' \
-mask './groupmask/cortex_be+tlrc' \
-num_glf 1 \
-glfLabel 1 Group -glfCode 1 'Group : 1*A & 1*B & 1*C & 1*D Movement :' \
-num_glt 6 \
-gltLabel 1 A-C -gltCode 1 'Group : 1*A -1*C Movement :' \
-gltLabel 2 B-C -gltCode 2 'Group : 1*B -1*C Movement :' \
-gltLabel 3 D-C -gltCode 3 'Group : 1*D -1*C Movement :' \
-gltLabel 4 A-B -gltCode 4 'Group : 1*A -1*B Movement :' \
-gltLabel 5 A-D -gltCode 5 'Group : 1*A -1*D Movement :' \
-gltLabel 6 B-D -gltCode 6 'Group : 1*B -1*D Movement :' \
-dataTable \
Subj Group Movement InputFile \
Subj001 D 0.0712 ./subj001/DC/DC_Z+tlrc \
...
2

Daniel,

When deciding on centering strategy, do you believe that those groups are intrinsically different in terms of movement? In addition, the following line

-qVarCenters ‘(0.0450,0.0607,0.0575,0.0492)’ \

is not valid. If you want to center each group separately, you have to manually perform the centering yourself before putting the quantitative variable values in the data table.

if I look at the results of the MVM on the brick showing only the effect F of “Group” variable
I don’t get exactly the same results as the ANOVA without the covariate

They should be different because including an extra explanatory variable in the model would change the dynamics of the model. Check out the Movement effect in the output and see if it’s reasonable to include the covariate.

is it the same to put the quantitative variable Movement inside glfcode and gltcode or not?

It’s not clear to me why you need the GLF specification. If you are interested in group comparisons, you should not include the “Movement :” in the GLT specifications. Do that only if you want to compare those groups about their Movement effect (slope).

3

Good morning Gang, this was helpful.
So, if I understood well, the group effect controlling for the movement variable is in the first sub-brick called “Group F”, in the resulting dataset , the sub-brick “Group:Movement F” is the group effect of group on the slope of movement, correct?
So the t-tests I have now with this code test the difference between the slopes of the movement between the groups? It would make sense this way (:smiley:
Thank you so much, sorry for the misinterpretation.

4

Daniel, the answer to all the three uestions of yours is yes.

5

Great, thank you again.
Daniel