Hello all,
I am currently working with a mixed design. We have overall alterations between rest periods of 30 seconds and then performance of a task for about 2 minutes. Within the task blocks we have jittered ISI and a general framework of a normal fast event-related design. Within this event-related part we are interested in back sorting based upon subsequent memory obtained outside of the scanner. Memory performance is right around 70% and we have 200 trials across two different runs.
There is some internal debate in the lab as to what EV's are needed in the design. We are settled that there should be one for the block effects (task vs. rest) as well as an EV for the remembered, and a third for forgotten trials. The discrepancy comes when determining if we need a fourth EV that codes for a generic "trial" condition. Such an EV would be a perfect linear combination of the remembered and forgotten trials EV's. The logic put forth for the inclusion of such an EV is that as the remembered condition has more trials (70%), more of the generic activity just associated with performing the task irrespective of memory performance will load on the remembered EV and come out when examining the difference between the remembered EV and forgotten EV.
What is the correct design matrix? Do I need to worry about the differences in trial numbers between the two conditions? Do I need to include this generic "trial" EV or is the correct design having 1 EV for the block effects, 1 for remembered trials, and 1 for forgotten trials? Any advice you all have on this would help clarify this issue for our group.
Brian Gordon
|