Thanks for your reply. In the end, the cause seems to have been that I had a switch flipped in one of my scripts:
# Carry out pre-stats processing?
set fmri(filtering_yn) 0 <- should be a 1
Curiously, it works both with 110 and 111 entries now. I am sticking with the 110 because my previous analyses have been run that way and the time course follows almost precisely the timeseries of a nearby voxel without appearing to be delayed by a volume. As well, the contrasts that come out seem to make sense and the negative versions don't. (Is there an additional method for testing this or understanding how FSL is lining the EV one-entry-per-volume up with the number of volumes including a deleted one at the front?)
|