So, I have a group of nations that exhibit global spatial dependence. I want to partition the nations into groups so that, at an intuitive level at least, the "sum of the spatial dependence within each group" or some similar measure is maximized. This problem seems similar to local statistics, but I'm not looking for "hot spots"; I want to partition all the nations. It also seems similar to spatial regimes, but its an issue of dependence, not heterogeneity.
Anyone have any suggestions?