Hi Everyone,
I have a blocked design data set with that I'm trying to model in SPM5.
Some of the VOIs in the data (amounting to ~6% of the voxels) have unusually
long response delays (BOLD response onset after 9-11 seconds after the
stimulus onset), although most other regions have the usual 4-6s delay.
So the question is how best to model this. I've been using canonical
HRF plus the two derivatives, but the derivatives don't seem to make a
difference at all (as judged by leaving them out of the model or the
contrasts). I thought the derivatives were designed to capture variations
like this. So my first question is how unusual this is (i.e., derivatives
not making a difference), so I know whether I should give up on the
derivatives or not.
Second, when I set the epoch onsets (at the model specification stage) to
reflect the 9-11s delay, the long-latency VOIs are beautifully modeled. So
my other question is whether it is somehow a bad idea to customize the
onsets this way (i.e., in this case using two different latencies, 6s and 10s).
Thank you very much,
Jay Hegdé
University of Minnesota
|