A Nonparametric Bootstrap Test for the Equality of Coefficients of Variation

Abstract

Abstract The problem of testing the equality of coefficients of variation for two different populations leads to various levels of difficulty depending on the possible assumptions on models and parameters for the two populations under study. Simulation techniques appear to be the only feasible way in the case where the available information came only from data and when these do not allow one to make any assumption on the models. In this work we propose a nonparametric bootstrap procedure, both to build the test statistic and also to approximate the p-value. The properties of the test and its critical aspects are illustrated and discussed by means of an application to a real data set of anthropometric measures for the study of the sexual dimorphism. Keywords: Ancillary testFrequentist p-valuePermutation testMathematics Subject Classification: 62G1062P10 Acknowledgments This research has been supported by M.I.U.R. (Ministero dell'Istruzione, dell'Univesità e della Ricerca), Italy. Notes (*) indicates that x and y could be assumed normal. Columns 2–3 show the means and variances of male (x) and female (y) measures.