Hi All,

I have a question regarding how to model my task. The fMRI task had an event-related design (optimized with OptSeq2), 20 trials of each of 4 experimental conditions and 20 nulls in each of two runs. It was a combined flanker and no-go task. In this paradigm, each trial is 3 seconds (two 1.5 second TRs) - starting with 500 ms fixation cross, 800 ms stimulus, and an additional 1700 ms blank screen allowing additional time to make a manual response (although the average RT is closer to 400 - 600 ms, which means they are responding within the first TR). THE CATCH, for me at least, is that the TR was 1.5 seconds, meaning that two complete scans were collected for every trial. So the judgment we are interested in is more or less occuring during the first TR of any trial.

When I started the experiment, I thought I was under good advisement that I could model the design by simply counting two TRs for any given trial BUT since then I have heard views to the contrary. I have done all the analyses counting both TRs, and I have re-done them only the first TR for each trial. So far, both approaches have yielded reasonable results but in either case they are quite different. Although the data look cleaner in the one TR approach, they (again although reasonable re: the task demands) do not really overlap with the findings from the two TR approach.

I am trying to figure out if there any logic to favor one over the other
approach (counting only the first or both TRs), including how this might
affect signal to noise and power.

Do you have any thoughts on what is the correct approach???
Thanks.
Brandon