Abstract

If W is a fixed, real-valued random variable, then there are simple and easily satisfied conditions under which the function dW, where dW(X,Y)= the probability that W “separates” the real-valued random variables X and Y, turns out to be a metric. The observation was suggested by work done in [1].