Hello FSL listers,

I'm trying to figure out why cutoffcalc is giving me such a long filter cutoff for my design. My data has 405 volumes (TR=0.72s). My design has 8 blocks (each block is a different EV) of 27.5s and 4 blocks of fixation (15s).
X-A-B-X-C-D-X-E-F-X-G-H

Basically, I tried to calculate the optimal HP cutoff according to 'cutoffcalc -t 0.95 --tr=0.72 -i design.mat', and received a filter period that was 1749.6s, which is 6 * (task duration). This didn't make any sense to me, since the worst case scenario should have been 2*(rAB+rCD+rEF+rGH).

For an alternate (shorter) task, I received a filter period of 90s (the minimum). So, I created a new design.mat using the EVs/contrasts from the alternate task but the image (duration) from the first task. I again received a filter period that was 1749.6s.

I ran a test of cutoffcalc, reducing the duration of the design step-by-step, but keeping the EVs from the alternate task. The filter duration stayed at 6*(task duration) until it sharply dropped and started yielding more reasonable filter durations - ultimately dropping back to 90s.

A few questions:
- Why does the upper-limit of cutoffcalc seem to be 6*(task duration), and under what conditions does cutoffcalc return that value?
- Do larger / longer designs pose a problem for cutoffcalc, or for HP filters in general?
- Does the issue arise from the fact that there is only one observation per condition? Does this pose a problem for estimation, or actually for using the HP filter?


Thanks,
Greg
____________________________________________________________________
Greg Burgess, Ph.D.
Staff Scientist, Human Connectome Project
Washington University School of Medicine
Department of Anatomy and Neurobiology
Email: [log in to unmask]