Hello AFNI team!
I have been trying to implement an amplitude-modulated GLM to account for response times in the model. For about half my subjects, everything seems to be running fine. But for the other half, a seemingly random selection of amplitude-modulated stimulus-regressors come out as all NANs (in the attached image, the red line is an example of this). Any ideas why the amplitude modulation would fail?
Never mind, I figured it out! When creating my stimulus timing files, I forgot to remove trials where the subjects didn’t respond. These trials had “nan” as the response time, which explains why 3dDeconvolve could not scale the regressor.
For that first condition’s timing file (the xmat regressors are from the bottom up), what is the output of:
timing_tool.py -multi_timing stimuli/TIMING.txt -multi_timing_to_event_list GE:ALL -
That should show the modulators being applied. If the input were stimuli/*.txt, then it would show all of them (though then we wouldn’t know which is the problematic one).
Great! Sorry for my late reply, though you might still be interested.
Thank you very much! I actually didn’t know about timing_tool.py and it looks like it will be useful to me in the future!