Robust estimation of change point in two-phase linear regression model

In this thesis, estimation of the change point in the two-phase linear regression model is considered. The change point in the two-phase linear regression model is estimated by using the modified maximum likelihood (MML) methodology, originated by Tiku (1967, 1968), and the one-step M (OSM) estimation methods under the assumption of long-tailed symmetric (LTS), generalized logistic (GL) and Jones and Faddy's (2003) skew t (JFST) error distributions. MML and OSM versions are developed for the methods proposed by Quandt (1958,1960) for discontinuous; Muggeo (2003) and Hudson (1966) for continuous models. The results of the Monte-Carlo simulation study demonstrate that the MML and the OSM estimators are more efficient and robust than least squares (LS) estimators. Furthermore, proposed methods are applied to the real life problems taken from the literature. They are shown to be more efficient than the corresponding LS estimators.