Dear Mike,
You shouldn't go with separate models. Either you can separate those trial components, then go with a corresponding model, or you cannot, then go with another model.
In your case, you can't separate TS-E from WM-E as WM-E always follows TS-E, and it follows instantaneously. However, you can contrast E and C (as you planned with GLM-2). I'm not sure how to interpret that contrast though. You say subjects didn't have to perform WM during C. Did they know so before the stimulus was presented, e.g. as visual cues differed? In that case the comparison E vs. C would not just correspond to stimulus and WM related activity vs. stimulus related activity only, but there would be a certain bias, as only during E they would have to process the stimulus actively, while during C, they might try to ignore it.
How to solve this issue would depend on the aims, e.g. 1) one could vary the length of the WM period (this might lead to a bias on each own though), 2) one could start with a visual cue indicating a WM trial, but after stimulus presentation one could present another cue that they can forget about the simulus.
Best regards
Helmut
|