Bootstrap comparison of statistics for testing the homoscedasticity of random fuzzy sets

Abstract

The problem of testing the equality of variances of k random fuzzy sets has been recently developed on the basis of Levene's classical procedure. Asymptotic and bootstrap approaches have been carried out in this framework, and the proposed test was compared with a Bartlett-type test. In this work, a deeper comparison between some bootstrap statistics based on both Levene's and Bartlett's classical procedures for testing the homoscedasticity of several random fuzzy sets is analyzed. The empirical behaviour of those statistics is investigated by means of simulation studies concerning both type I and type II errors. © 2013 Springer-Verlag.