As I understand tent functions, they allow for stimulus onset times that aren’t time-locked to the TR by assigning an event to two TRs proportionally, so that the peak of the tent falls at the onset time. I’m plotting a design matrix and I notice that for the stimuli that aren’t time-locked to the TR, the peak is smaller than for those that are. I always assumed that if (for example) the TR is 2 seconds and the stimulus onset time is 3, the first TR would get assigned .5 and the second TR .5, with a peak of 1 at 3 seconds. Is the fact that the peak < 1 an artifact of plotting, with something going on “under the hood” with 3dDeconvolve that makes the peak 1 at 3 seconds that doesn’t show up in the plot, or am I misunderstanding how tent functions work?
You should not see the peak in any regressor off the TR-grid.
In your example, with a desired peak at time=3s, as you say,
the adjacent regressors would shows values of 0.5 at time=4s
and time=6s (and onward). But there is no regressor or data
at time=3s (or 5s, etc.), to see the actual peak in.
For events that occur on the TR grid, values of 1 should be seen.
For events that occur off the TR grid, values of 1 should not be seen
(where the 1.0 is shared between adjacent regressors, say).