To estimate probability distributions, you need data that is non-random. Non-random means there's a pattern. That's another word for correlation.

In using those methods you may never calculate correlation as a number, but when those methods find something you still rely on the fact that there is a correlation.