Ramadha Piyadi Gamage

Empirical Likelihood (EL) method introduced byOwen (1988) is a widely used non-parametric tool for construct-ing confidence regions due to its appealing asymptotic distri-bution of the likelihood-ratio-type statistic which is same as theone under the parametric settings. However, the EL method wasintroduced to be used with independent data, hence it becomesdifficult to apply it to dependent data such as time series mod-els. An EL-based detecting procedure for structural changes intime series data is developed. The asymptotic results such asthe asymptotic null distribution of the test statistic, the consis-tency of the estimator of the change location and the test areestablished. Further, simulations are conducted to investigatethe performance of the proposed test under various scenarios.Additionally, real data is used to illustrate the detection proce-dure.