Dear SPMers,
In an event-related task, I create an A >> B contrast. "A" always contains
44 events, but "B" varies. If B happens to contain just a few events (like 3
or 4) my signal decreases and I get less or no activation. On the other
hand, when B is like 20, then I start to get something. My question is: in a
group analysis (second order), should I set a threshold and include only
subjects for whom B was big? If I include everybody, are the small-B
subjects going to decrease the group overall activation?
Thank you,
Daniel
|