A test of homogeneity of scale parameters based on Gini's mean difference

Abstract

In this article, a multi-sample test for homogeneity of scale parameters against simple ordered alternative based on Gini's mean difference is proposed. The critical points through simulation are computed for the proposed test in case of standard exponential, standard logistic and standard uniform distributions; however the proposed test can be applied even in case of other distributions like Laplace, Pareto, Weibull etc. Construction of simultaneous one-sided confidence intervals (SOCIs) is proposed along with an illustration. Power of the proposed test is computed and some power comparisons are also made. Statistical simulation revealed that the power of the proposed test relative to some existing tests is higher.