Testing variability orderings by using Gini’s mean differences

In this paper, we derive a measure of discrepancy based on the Gini’s mean difference to test the null hypothesis that two random variables, which are ordered in a variability-type stochastic order, are equally dispersive versus the alternative that one strictly dominates the other. We describe the test, evaluate its performance under a variety of situations and illustrate the procedure with an example using log returns of real data.