To AFNI-ers,

We are working on defining ROIs using 3dmerge and have run into a snag:

3dmerge allows one to specify how clusters are arranged with options such as "1clust_amax, 1clust_order, etc.".

however, I've found that when using 1clust_order, the program ignores the SIGN (+/-) of the underlying data which leads to two very distinct areas of activity being grouped as one.

To be more specific, we are doing a comparison of Category A vs. Category B. These categories activate adjacent regions in temporal cortex. The output of my multiple regression yields nice negative correlations for CatA and positive for CatB. When I use 3dmerge to separate the two clusters it works ONLY if I use 1clust. However, this picks up ALL clusters that are active. I would like to use 1clust_order so I can select each cluster independently. But when i do, the program ignores the sign and groups the 2 clusters as 1, eliminating the interesting contrast.

Is there an option I am unaware of that will allow 1clust_order to pay attention to the sign and NOT group these two clusters together?

Thanks!

A