Hi Mike,
The cluster method that you should employ with 3dClustSim
is simply the one that you want to use. We do not know.
The difference between NN1, NN2 and NN3 is whether neighbor
voxels should be considered as part of the same cluster.
NN1 - requires faces to touch
NN2 - faces OR edges
NN3 - faces OR edges OR corners
So NN1 will generally have the smallest clusters, while
NN3 will have the largest (because some of the smaller
ones might group into a larger one).
We tend to focus on NN1, but that is up to you.
Similarly, whether to use 1-sided, 2-sided or bi-sided
is up to you. For example, consider a t-test of A-B.
Do you want to only positive results (where A > B)? If
so, that is a 1-sided test.
Otherwise, do you want the positives and negatives to
cluster together? If so, that is a 2-sided test.
Otherwise, you want both positives and negatives, but
they should not cluster together. That is a bi-sided
test.
- rick