>let me see if I understood. you suggest me to replace my 6 EV (2 EV for the three cong/incong/neutral conditions) with just 2 EV.
>that is, not consider the trial type but just put an EV with fixed 3rd column value (1) and one with variable values (xRT-baselineRT)
>and then create one simple contrast [0 1].
In general yes.
>if so, consider that i have two types of conflict (defined by the 2 possible "TARGET stimulus" characteristics), that is 2 cong 2 incong and 2 neutral, and one of my question is why/where they behave differently (as i see from RT). do I have to add two further regressors or just the conflict-related (with its different xRT-baselineRT) one?
That depends. If you believe that there are two independent cognitive processes (conflict1 and conflict2), then you should model them as separate variables. In general, you should have as many regressors as there are temporally distinct cognitive or neural processes.
>and make [0 1 0]/[0 1 0 0] and [0 0 1]/[0 0 0 1] to assess the two, presumably different, conflicting voxels patterns?
yes. a contrast for each regressor that represents a cognitive process.
>since in all conditions they have to finally press a button answering to a question regarding the "CUE stimulus" characteristic (which is independent from the conflict type), should i add a further EV in order to let him explain part of the variance.
You can certainly add a motor response regressor which is an impulse positioned on the response. Just remember that your regressors may be highly correlated.
cheers,
jack
|