Robust Testing between Two Asymmetric Distributions: a Simulation Study

In this paper a robust test statistic for testing between two income distributions is analysed. Victoria-Feser (1996) proposed a robust version of the classical test based on the comparison of the maximised likelihood functions for the two models (Cox 1961, 1962). She showed its robustness properties as well as other properties. She also proposed an algorithm for the computation of the robust Cox-type test statistic. This algorithm is initialised by taking the classical maximum likelihood estimates of the parameters. But would it not be more appropriate to take a robust estimator to estimate the parameters of the models? The aim of this paper consists in answering this question in use of simulated examples. S-Plus functions for computing the described procedures are made available.