No rest in between trials with RSFgen?

AFNI Message Board
1

In my experimental design, there are two trial types and we want to time-lock the stimulus to the TR. To create the stimulus schedule, I am running RSFgen in a similar way as described in: https://afni.nimh.nih.gov/pub/dist/HOWTO/howto/ht03_stim/html/AFNI_howto.html. I ran RSFgen specifying 50 ‘A’ trials, 50 ‘B’ trials, and 100 ‘rest trials’.

When using RSFgen in this way, there are often times when two (or three, or four) trials end up being back-to-back with no rest trials between them; in effect, you occasionally get an ISI=0. Is this not problematic from the perspective of deconvolving the HRF if you end up having trials that are back-to-back like that on occasion? Thank you for your help in clarifying this matter.

2

You should use the newer program make_random_timing.py which has more options to control the timing.

3

… and adding to Bob’s comment-- you might be wondering, “Well, how should I know if an AFNI program is old or new?”

The (only very recently available) answer is that there is a bit of guidance online:
https://afni.nimh.nih.gov/pub/dist/doc/htmldoc/educational/classified_progs.html#make-edit-evaluate-stimulus-timing-files
(and actually, it is still a work in progress as we continue classifying things further.)

The above link might be useful when you are searching for a tool in AFNI to use for a particular task.

–pt

4

Okay, thanks for the suggestion to switch to make_random_timing.py. I’m still confused about the question I originally asked you though. Generating the stimulus schedule with make_random_timing.py results in stimulus schedules where some trials are back-to-back and with no rest in between them. Looking at one of my schedules, there is even the case where there are 4 trials that end up being back-to-back with no rest in between them. Is this problematic from the perspective of deconvolving the HRF if you end up having trials that are back-to-back like that somewhat often? I am running several iterations of this routine, testing with 3dDeconvolve to get the efficiency for the contrast I’m interested in, and then taking the most efficient schedule, but I’m still wondering if I need to do additional QC of the stim schedules by excluding any schedules where there is too high of a proportion of ‘back-to-back’ trials, or too many back-to-back trials in a row?

For comparison’s sake I also tried testing out make_random_timing.py without time-locking it to a TR and there are also instances where there is no rest between trails (effectively an ISI=0) but it was very rare compared to creating stimulus schedules with the time-locking.

Thanks so much for your conceptual help with this!!

5

With the basic usage of make_random_timing.py, there
is a -min_rest option. With the advanced usage, one can
provide min/mean/max rest restrictions, along with choices
for the probability density function that correspond to both
the stimulus timing and the corresponding rest (ISI).

Please see the -help output: make_random_timing -help

  • rick