Hello AFNI team,
I ran 3dREMLfit on my single subjects, to then use the coefficients and t-stats with 3dMEMA (paired contrast). The output of 3dMEMA was then used to define a mask within which I wanted to perform a statistical test that looks on how the contrast is modulated by different covariates.
To do that, within the mask, I ran 3dMVM on the coefficients obtained with 3dREMLfit, and defined my desired GLTs.
Is this legit? Or must I use 3dMEMA on coefficients obtained with 3dREMLfit? If so - how do I define GLTs with 3dMEMA?
Also, any tips on the advatages of 3dMEMA over 3dttest++ would be appreciated (I chose 3dMEMA just because the results looked a little better compared to what I got with 3dttest++, no strong theoretical reasoning).
Hello AFNI team,
I ran 3dMVM on the coefficients obtained with 3dREMLfit, and defined my desired GLTs. Is this legit?
Yes, that’s fine.
any tips on the advatages of 3dMEMA over 3dttest++ would be appreciated
Ideally the analysis for a dataset should be performed with one integrative model. In real practice this is not always feasible in neuroimaging due to a couple of limitations: model complexity, computational cost and the need for data quality control. Therefore, the analysis pipeline is usually split into two steps: one for each individual subject, and another one for group analysis. Such a two-step approach may result in information loss since the group analysis only uses the effect estimates (betas) as input most of the time. One way to make up for the information loss is to take the uncertainty of the effect estimates (embedded in the t-statistic) into consideration during group analysis. This pretty much explains what you experienced in your analysis (“I chose 3dMEMA just because the results looked a little better compared to what I got with 3dttest++”).