I have read about it in testing materials when discussing project management planning. I believe it is simply taking the historical information based on the number of bugs discovered in X component, then using those metrics to determine the amount of time that X component should have allocated to testing.
I am keeping track of the number of bugs I find for a certain class of inputs. If I find more bugs with a certain class than others I can prioritize and even eliminate test cases.
Of course this is just my interpretation of "statistical testing" techniques.