> I actually tried the -stim_nptr option first. Since I have one image every 1.5
> TR, I used the following set-up: -stim_nptr 1.5
I believe the option should be something like
-stim_nptr k 2
where k is the index for stimulus. The reason is that each event in your design lasts for 3 seconds, e.g., 1.5 TR's. So you want to break each TR into two finer grids to adapt to the situation. And that is why you want a number of 2 in the above option. You have to be very careful with other options with this solution. Read 3dDeconvolve manual for more details on page 23:
http://afni.nimh.nih.gov/pub/dist/doc/manual/3dDeconvolve.pdf
Gang