Dear Christian,
Thanks again for an answer, now at least I know what you mean by sensible
histogram. It seems however that your answer assumes that the situation I am
describing must result from some kind of error. I am not that sure about it.
I have got 40 subjects in the analisys, in simple visual blocked design so I
guess when simple effects are considered one would expect largely
significant results.
Anyway lets assume for a while that my results are OK, and lets reiterate
my original question, namely how can you avoid getting one cluster spanning
through entire brain? Or how can you avoid getting cluster covering two
distinct brain structures. Quite often you get one cluster when two big
patches of activity are "connected" by a narrow lets say bridge of active
voxels. Is there any way of separating them? In other words is there any
utility available to manipulate the span of clusters? Any measure of
closeness ? I tried the cluster command but without much success. I thought
that maybe the --connectivity switch governs the span of the cluster but it
proved wrong (changes to connectivity value did not affect the span of the
clusters at all). There must be some algorithm that decides where one
cluster ends and another one begins. Is there any way of controlling how
this algorithm works?
Best regards
Mike
|