Hi Aaron,
We would generally recommend that you clean up the within-subject timeseries for motion by using a set of motion regressors and/or outlier cleanup and/or ICA denoising (e.g. with FIX). In addition it is fine to also include a mean FD regressor for the between-subject analysis in the GLM (as this might protect you a bit more against motion effects), but if you are concerned about motion then the most effective thing will be the first level cleanup.
All the best,
Mark
> On 11 Jul 2017, at 21:56, Aaron Tan <[log in to unmask]> wrote:
>
> Hi all,
>
> Is it appropriate to using meanfd as a nuisance covariate in my GLM for resting-state data? I am using randomise on seed-based correlation maps. I imagine it might control for variance that is a result of certain regions having a greater likelihood of producing artifactual correlations due to motion.
>
> Thanks a lot,
> Aaron Tan
|