Comment by lukego

Is the question fundamentally: what's the relative likelihood of each number or clusters?

If so then estimating the marginal likelihood of each one and comparing them seems pretty reasonable?

(I mean in the sense of Jaynes chapter 20.)