Performance of Methods for Testing the Ratio of Coefficients of Variation of Two Gamma Distributions

Abstract

This paper presents two methods for testing the ratio of coefficients of variation of two gamma distributions. The proposed statistical tests were based on the methods of score and the Wald interval following the concept of the method of variance of estimate recovery (MOVER). To compare the performance of the proposed statistical tests, a simulation study was conducted under several shape parameter values in two gamma distributions. The performances of the statistical tests were compared based on the empirical size and the empirical power of test. The simulation results showed that the statistical test based on the method of score interval performed better than the statistical test based on the method of the Wald interval in terms of the nominal significance level attained and empirical power of the test, and is thus recommended for researchers. The performances of the proposed statistical tests also were examined through a numerical example.