Hi Paul,
Thank you very much for your reply.
As indicated here (https://afni.nimh.nih.gov/afni/community/board/read.php?1,161034,161745#msg-161745) when doing what you've suggested I got an incredibly high number of motion censored volumes - even though the motion is not apparent in the original EPI files for each run which was very surprising for me. Hence, I tried to disentangle the effects each step of the pre-processing would have on my data (https://afni.nimh.nih.gov/afni/community/board/read.php?1,161034,161922#msg-161922), especially using the quality check html file. I'm very new to pre-processing in AFNI, but from my layman point of view, it seems that detrending first as suggested in your initial answer (https://afni.nimh.nih.gov/afni/community/board/read.php?1,161034,161061#msg-161061) imposes problems. Likewise does the demeaning. To circumvent the problems of motion censoring if there is de facto no motion (as observed in the original EPI), I thought it might be better to do the pre-processing firstly and do the concatenation/demeaning as a separate step.
This also has the additional benefit that it is easier to select different video clips to be included for the ISC calculation. I study memory, so I'm interested in comparing the ISC for remembered video clips (i.e. both members of a given pair have remembered) with the one for forgotten video clips (i.e. both members of a given pair have forgotten). However, this is different for each pair of subjects, so the concatenation has to be done individually (i.e. 2(memory outcome: remembered vs forgotten) * 1/2*N*(N-1) = 2450 concatenations given my N = 50) and for me, it seemed easier/more efficient to pre-process the data of 50 participants to then use the 50 *errts files to create those 2450 final time series by concatenation than to first concatenate and then pre-process 2450 time series.
Best regards,
Stef
