Amplitude Modulation with Missing Values

Hi AFNI experts!

I’m running an amplitude modulation analysis with a trial-by-trial behavioral response as my AM. Therefore, my timing files look like this: onset*behavior:duration. However, for some trials, the participant didn’t give a behavioral response. So I’m unsure what to do with those trials. Do I delete them from the timing file (but then I’m not sure if I’m modeling the implicit baseline correctly). Or, is there a way to just have a missing value there so I can still model those trials, but without the AM. Here is an example of a timing file with missing values (I have them currently coded as 99 just so I can visualize them).

63.2883:3 111.0543:3 135.68899:3 146.2541:3 156.3214:3 195.0212:3 260.9214:3 273.9885:3
28.5764:3 40.1432:3 99.4764:3 135.6765:3 156.30999:3 249.3425:3 260.9094:3 286.0425:3
165:3 28.5675:3 75.8342:3 122.13:3 228.72:3 249.3333:3 260.94:3 273.9674:3

Thank you!

There might be 3 reasonable possibilities:

  1. Move the trials without a modulator to a different regressor class.
  2. Assign an average response time to such trials (so they do not affect the modulation beta).
  3. Assign a maximum (but reasonable) response time to such trials (because that seems more natural based on the task).

Which to do depends on the task and the circumstances of a failed behavioral response. Was the subject trying but just too slow? What the subject not paying attention?

  • rick

Thanks so much Rick! To give more information, the behavior we’re looking at is actually aggressive behavior. In the task, the subject chooses the volume level of a white noise blast to “send to other participants”. This is on a scale of 1 (no noise) to 5 (max noise) and is our measure of aggression. We’re going to look at aggression to mean vs nice peers (in a contrast). Because of this, I’m not sure if the third option is the best choice. However, the 1st or 2nd option seems more reasonable. To move the trials without a modulator to a different regressor class, would this then be a whole separate timing file? If so, I’m guessing I would then treat that event as “noise” and not use it in my contrasts?

Thank you!

Sorry one other note, the circumstances of a failed response could be they didn’t respond in time, they weren’t paying attention, or they simply didn’t want to respond to that trial (even though they were supposed to respond to every trial). So it’s hard to say!

Yes, it seems like making a separate regressor might be the way to go.
Yes, that would mean putting those events in a separate timing file.

The other difficulty here would be in addressing whether some subjects have no such events.
In that case, you might still prefer to have that (possibly empty) timing file for all subjects, just to be consistent with the regression output. But that would also mean adding -GOFORIT options, to tell 3dDeconvolve/3dREMLfit to proceed, even with bad-looking input.

Without all subjects having the non-AM event

This gets slightly messy…

  • rick

Ok thank you so much, Rick. I’ll let you know if I have any further questions!