Hi Paul,
Thank you for your reply. I tried to explore where in my pre-processing the high degree of censoring occurs. From how I understood your reply from April 12th, I firstly detrended each run, then took each each clip, demeaned it, unshuffled it and combined it into one nifti file which I then used as input for the pre-processing following the same code as used in Chen et al (2016, NI) with the modification of setting -regress_polort 0. Even though the raw runs didn't show any severe motion by visual inspection (and I also verified this by using the raw files as input for the preprocessing, which for an example subject then had a censor fraction of 0.0018); however, if I follow the steps as described, I get a censor fraction of 0.8381 for the same subject.
I then tried to disentangle these strange results: as said before, if I just enter all three runs as they are, I barely have censored TRs. If I detrend them firstly (using 3dDetrend -polort 2; separate for each run), the censor fraction increases to 0.5145. If I combine all three runs into one run (without detrending in all following steps), just the step of combining them into a big run slightly increases the censor fraction to 0.0299 and the censored TRs are at the two intersections between run 1 and run 2 and run 2 and run 3, respectively. The same issue also occurred when I further concatenate the single clips into one run without unshuffling them (censor fraction 0.0337) and again, these censored TRs are where the runs are combined. If I then add the unshuffling to it, the censored TRs distribute (censor fraction 0.0236), but are at the beginning of a clip and the mot enorm plot shows several spikes aligning with the begin of a video clip. If I demean the video clips firstly before unshuffling them, this increases the censor fraction to 0.8145.
I am not sure on how to deal with this, as both, detrending and demeaning seem to impose severe problems for the later preprocessing. What would be your advice on this?
Again, many thanks for your support.
Best regards,
Stef
