Dear Afni experts,

We have a question regarding the interpretation of values after running a MVM.

The model used the beta values from GLM maps. the MVM analysis is based on a 2x2x3 model where there are 2 groups (OMT - SHAM), 2 tasks (Heart - Sound) and 3 time points (t0 - t1 - t2) and additional covariates, i.e. sex and age. So the MVM model is the following:

bs: Groups*age*Sex

ws: task*timepoint

qvar: age

We decided to run, then, different post-hocs combining different groups, time points and tasks. An example is:

- gltlabel 1 OMTvsSHAM -gltCode 1 ‘Group : 1
*OMT -1*SHAM time : 1*t1 -1*t0 task : 1*Heart’

In afni the structure of the output is as follows:

#0 (intercept) F

#1 Group F

#2 time F

#3 group:time F

#4 OMTvsSHAM

#5 OMTvsSHAM t

…

After getting the MVM results and maps, we double-checked the output #4. Are these the mean beta values?

Then we would like to verify if those values correspond to the mean of beta values computed by GLM. the latter mean values were calculated according to groups, tasks and timepoint. A bucket has been created with n maps. Each map represents the mean beta values per group/task/timepoint.

Now we selected a voxel from #4 of the MVM and found the value. Then we manually extracted the beta values from GLM maps and applied the model #4. The results we manually got are different from the one coming from MVM analysis. Is there an explanation? is there anything we are missing?

Thank you in advance for your help.

Regards

Piero