Dear AFNI experts,

During my experiment, participants watch 36 short movie clips (average duration 32 seconds) over three runs (i.e. 12 clips in each run) and are asked to give ratings after each clip. The order of movie clips is pseudo-randomised across participants and runs.

I am planning to follow the preprocessing as described by Chen et al. (2016): [afni.nimh.nih.gov] and have some questions about it:

If I understand the description of the "-regress_apply_mot_types demean deriv" option correctly, this adds the demeaned motion regressors as well as the derivatives to the regression model. Is it correct that the derivatives model the potential scanner drift over time so that this is accounted for? Further, is it correct that only the motion parameters are demeaned, but not the BOLD time series itself?

After the preprocessing of the three runs, I combine them and then remove all volumes associated with fixation between the clips and ratings participants are asked to give after each clip. Additionally, the volumes are re-ordered so that the resulting time course reflects the same order of movie clip presentation across participants. This then results in a concatenated time series only associated with the display of the clips itself which is later used in the 3dTcorrelate command. I think that for computing the intersubject correlation, I should the option "-polort -1" given that my data is preprocessed and concatenated. Is that correct? But this assumes that the data is detrended within the preprocessing, so presumably with the "-regress_apply_mot_types demean deriv" option?

Given that I concatenate my data, would it be recommendable to demean the time course associated with each clip (using the mean of the volumes associated with each clip) before merging the volumes for all clips together?

Lastly, should I include bandpass filtering to the preprocessing? If so, at what stage would it be best implemented?


I am looking forward to your thoughts on the issues raised above, any help is highly appreciated!

Many thanks,
Stef