3dMVM input files question

AFNI Message Board
1

Hello,

I am trying to use 3dMVM to compare 2 groups with two crossed conditions (at least 6, up to 9) within each subject (bsVars A, wsVars A*B). The data were acquired in an event-related rather than blocked manner, so in other words the conditions are intermixed. I have timing files (that I used with regress in afni_proc.py) for each condition for each subject, but currently the sub-briks for each subject are just each TR.

I know these questions will seem very basic, but I want to clarify:

  1. To do this analysis, I think I cannot use 3dANOVA3 because then I would not be able to model the subject-level effects (as random effects). Is this correct?
  2. Do need to (and if so, how can it be done) make new briks for each subject that correspond to each combination of conditions? I feel like this must have been done as part of the previous analysis I did (regress). But if not, how do I use the timing files? I think I must use 3dDeconvolve somehow?

Thanks in advance for your help.

2

compare 2 groups with two crossed conditions (at least 6, up to 9) within each subject

So, you have two groups of subjects and a within-subject factor? What do you mean by “at least 6, up to 9”? Does the within-subject factor vary in terms of the number of levels across subjects?

3

Sorry about that. Let me clarify.

The number of within-subjects factors does not vary across subjects or groups. There are two within-subjects factors (conditions) with 3 levels each, so a total of 9 combinations. (in one version of the analysis, we may omit one of the levels for one of the factors, so in that case it would be 2*3 = 6 combinations)

Thanks.

4

I think I cannot use 3dANOVA3 because then I would not be able to model the subject-level effects (as random effects).

Use 3dMVM.

Do need to (and if so, how can it be done) make new briks for each subject that correspond to each combination of conditions?
I feel like this must have been done as part of the previous analysis I did (regress). But if not, how do I use the timing files? I
think I must use 3dDeconvolve somehow?

Each combination of conditionsis the corresponding regression coefficient (beta) in your 3dDeconvolve/3dREMLfit ouput.