Estimation of $K$ (the number of populations)

In our paper describing this program, we pointed out that this issue should be treated with care for two reasons: (1) it is computationally difficult to obtain accurate estimates of ${\rm Pr}(X\vert K)$, and our method merely provides an ad hoc approximation, and (2) the biological interpretation of K may not be straightforward. While these caveats should be taken seriously, I should point out that in our experience the method usually gives sensible results, provided that appropriate care is taken.