Applying MBA approach to Connectivity-behavior analysis

AFNI Message Board
1

Greetings,

I'm new to BRMS but interested in tweaking the code underlying MBA to run some type of Connectivity-behavior analysis. The idea would be to look at how for example changes in connectivity between two conditions relates to changes in some behavioral measure.

Assuming the following default code:

brm(deltaY ~ 1 + (1|Subj) + (1|ROI1:ROI2) + (1|mm(ROI1, ROI2, weights = cbind(w, w), scale=FALSE)) + (1|mm(ROI1:Subj, ROI2:Subj, weights = cbind(w, w), scale=FALSE)) ...

one can control for covariates by running I believe:

brm(deltaY ~ 1 + Covar + (1|Subj) + (1|ROI1:ROI2) + (1|mm(ROI1, ROI2, weights = cbind(w, w), scale=FALSE)) + (1|mm(ROI1:Subj, ROI2:Subj, weights = cbind(w, w), scale=FALSE)) ...

however if purposefully want to investigate behavioral correlates should I run:

brm(deltaY ~ 1 + (1|Subj) + (1|ROI1:ROI2:**deltaBehavior**) + (1|mm(ROI1, ROI2, weights = cbind(w, w), scale=FALSE)) + (1|mm(ROI1:Subj, ROI2:Subj, weights = cbind(w, w), scale=FALSE)) ...

if so, would I then just calculate the Average Marginal Effect of deltaBehavior predicting deltaY at each ROI1:RO2 pairing?

2

@sinandrei Maybe something like the following?

brm(deltaY ~ Covar + (1|Subj) + (Covar|ROI1:ROI2) + (Covar|mm(ROI1, ROI2, weights = cbind(w, w), scale=FALSE)) + (Covar|mm(ROI1:Subj, ROI2:Subj, weights = cbind(w, w), scale=FALSE)) ...

Upon further contemplation, I find myself increasingly cautious about relying solely on correlation-based analysis. Consequently, the potential productivity of this approach remains uncertain to me.

Gang Chen

3

Thank you! My primary interest at the moment just lies in understanding how some of the traditional analyses look like in the rba-mba-brm world.