p-values, hypothesis testing and Ising model detection, Real world county datasets

A lot of researchers seem to be using Ising priors on the context they they help achieve better error rates and hence must be informative.
Can the p-value approach be better? Yes: But requires partition function calculation or atleast tight upper bounds. TRBP looks like it will get the job done. Well, at least in most cases.

Here is a compilation of all county-datasets in US.