I have some question about motion parameters for the regression analysis after processing my data through fMRIprep and TEDANA.
Are all 6 head motion parameters ( trans_x , trans_y , trans_z , rot_x , rot_y , rot_z) from the confounds_timeseries.tsv on the same scale? Would the rotations be on radians or in degrees? I was unsure if I would need to convert any numbers before running the regression analysis.
Is there a recommendation on how I would go about getting the censor files for the glm? Currently, I am planning on using AFNI’s glm analysis script which has censoring motion as one of its option.
Lastly, would I still need to make a censor file since TEDANA is run?
Thank you so much! I do apologize if these questions were already asked or the information is out there.
I think you’re asking about putting motion regressors into tedana as external regressors? We are still working on validating best practices. Within tedana, we’re just fitting all the time series so scaling and radians vs degrees should not matter. A scaled time series would have a different fit magnitude estimate, but the R2, F, and p statistics should be the same.
tedana does not currently handle censoring, although it is something we’re discussing. (See Censoring data within tedana · Issue #1053 · ME-ICA/tedana · GitHub ). If you have metrics from AFNI or fMRIprep that recommend censoring volumes, then I’d still censor them after tedana.
FWIW, the translation parameters are in mm and the rotation parameters are in radians, although, as @handwerkerd said, the scale won’t matter for the decision tree. The motion parameter units are in the confound file’s JSON in recent versions of fMRIPrep.