I am working on some functional analyses for which I have correlated voxel-based whole-brain activation with our behavior of interest across all participants using 3dTcorr1D. We would like to test for voxel-based group differences between younger and older adults on the relationship between our behavioral measure and brain activation (i.e., test for a significant group moderation effect). What would be the best statistical function and function options to use within AFNI to accomplish this goal?
Assuming that the alignment between two groups of subjects is reasonably fine, you can use 3dttest++ or 3dMVM for the group analysis you’re referring to. 3dTcorr1D is usually adopted for EPI time series analysis at the individual subject level.
After reading the documentation for both commands I want to make sure I’m accurately conceptualizing how I would run these analyses. If you could let me know whether my thought process about this is accurate that would be great.
With 3dttest++, for example, would essentially run a t-test to compare the younger and older adult groups to one another by putting the YAs into setA and the OAs into setB? Additionally, I know 3dMVM is used to perform an ANOVA with unequal group sizes. Can I use 3dttest++ with unequal group sizes, or would I need to run 3dMVM in that case as well?
If I were to use 3dMVM how would you suggest I set up the analyses? Would it just need one glt_code, such as
-gltLabel 1 OAvsYA -gltCode ‘age : 1Older -1Younger’
to look at the change of this brain-behavior relationship across levels of the age factor?
Please let me know if you need additional information or clarification for my questions.
With 3dttest++, for example, would essentially run a t-test to compare the younger and older adult groups to
one another by putting the YAs into setA and the OAs into setB?
That’s correct. Use option -covariates in 3dttest++ to specify the behavior measure as a covariate.
If I were to use 3dMVM how would you suggest I set up the analyses? Would it just need one glt_code, such as
-gltLabel 1 OAvsYA -gltCode ‘age : 1Older -1Younger’ \
to look at the change of this brain-behavior relationship across levels of the age factor?
Depending on your research hypothesis, you want to something like the following:
3dMVM -prefix … -jobs …
-bsVars "groupbehavior"
-qVars “behavior”
-gltLabel 1 OAvsYA1 -gltCode 1 'age : 1Older -1Younger’
-gltLabel 2 OAvsYA2 -gltCode 2 'age : 1Older -1*Younger behavior :’
…
Thank you for your previous help! I have a couple of follow up questions for you. I am performing the analyses we discussed above for the moderation using 3dttest++ and looking at option to do multiple comparisons correction using ClustSim and ETAC. The current documentation states that ETAC correction is not accurate for covariates. Is this still the case? Can the ETAC maps only be applied to the mean of the output and not the slopes?
If this is the case for ETAC, is this rule the same for ClustSim output? Or can the ClustSim output be applied to the cluster sizes of the covariate maps as well?
We are also looking into 3dMVM to perform a task by age ANOVA. Do ClustSim and ETAC work for 3dMVM analyses or do we need to do something to correct for multiple comparisons in those outcomes?
The current documentation states that ETAC correction is not accurate for covariates. Is this
still the case? Can the ETAC maps only be applied to the mean of the output and not the slopes?
I suspect that remains to be the case.
If this is the case for ETAC, is this rule the same for ClustSim output? Or can the ClustSim
output be applied to the cluster sizes of the covariate maps as well?
No, that’s only true for ETAC, not for Monte Carlo simulations with 3dClustSim. In other words, 3dClustSim works for any kind effect.
Do ClustSim and ETAC work for 3dMVM analyses or do we need to do something to correct
for multiple comparisons in those outcomes?
ETAC currently works only in 3dttest++. So for any other programs such as 3dMVM, use 3dClustSim.
The
National Institute of Mental Health (NIMH) is part of the National Institutes of
Health (NIH), a component of the U.S. Department of Health and Human
Services.