Empirical likelihood ratio test for the change-point problem

Abstract

A nonparametric method based on the empirical likelihood is proposed to detect the change-point from a sequence of independent random variables. The empirical likelihood ratio test statistic is proved to have the same limit null distribution as that with classical parametric likelihood. Under some mild conditions, the maximum empirical likelihood estimator of change-point is also shown to be consistent. The simulation results demonstrate the sensitivity and robustness of the proposed approach. A famous real example is studied to illustrate its effectiveness.