Dear Donald,
Thanks for your reply.
Indeed on the first level I have all 9 conditions modeled, but if I want to fit everything into one second level analysis model I run into a problem due to a quality difference in my switch/repeat factor. In my design I have 3 types of cues signalling 3 types of trials, where 2 of them are task trials (t1 and t2) and the 3rd - 'rest' trials. Thus my switch factor encompasses switches between tasks (t1 and t2) and also switches between task trials and 'rest'(waiting) trials, and that makes results hard to interpret. Having trouble figuring out a way to fit everything into a one flex factorial model, I started thinking of splitting it into a few, with switch/repeat factor (2 levels) and group factor (2 levels). For instance, 'rest' repeat trials (n-1 trial ='rest') and 'rest' after task1/2 trials (n-1 trial = task1/2), then the switch/repeat factor encompasses a repeat of 'rest' or a switch to 'rest' and thus there are no qualitative differences. Again the same with task trials (switch/repeat), not including 'rest' trials.
On the other hand, maybe I should already compute the contrasts of interest ( e.g. 'rest' repeat > task 'rest') on single subject level and do 2 sample t test on the second level?
Or maybe it is still possible to fit all into one model and in the end have an interpretable result?
Greetings,
Juste
|