Hello

Chapter 5 High Breakdown Unit Root Tests
In this chapter I suggest another selection to the OLS based unit cool off test of Dickey and Fuller (1979). The test departs from the genius suggested in Chapter 4 in that the present test kitty cope with a larger number of outliers. The behavior of this alternative test is studied using simulated data as well as the fourteen economic time series considered by Nelson and Plosser (1982) and extended by Schotman and van Dijk (1991a). The chapter largely draws from the sensible presented in Lucas (1995a). The setup is as follows. instalment 5.1 introduces the problem and motivates the extract of high breakdown point (HBP) estimators for testing the unit showtime hypothesis. air division 5.2 discusses the outlier mechanism and the MM estimator that is used. A introductory asymptotic analysis of unit root tests based on M estimators can be found in Section 5.3. A full discussion of the appropriate asymptotics can be found in Chapter 6. Section 5.4 compares the performance of the HBP unit root test with that of the standard Dickey-Fuller test by means of simulations. Section 5.5 presents the results of the robust and nonrobust unit root tests for an empirical data set, viz. the extended Nelson-Plosser series. Section 5.6 concludes this chapter.

5.



1 Introduction
In Chapter 4 I learn discussed the outlier sensitivity of the standard DickeyFuller t-test (DF-t) for a unit root. The solution proposed in that chapter was to replace the OLS estimator in the Dickey-Fuller procedure by an M estimator, in particular, by a pseudo maximum likeliness estimator based on the Student t distribution. This procedure went some way in making the DF-t less naked as a jaybird to anomalous observations. It is, however, well known in the robustness literary productions that M estimators can only cope with a hold in number of outliers. In the i.i.d. regression setting, one extreme outlier is fair to middling to corrupt the results obtained with an M estimator (see Hampel et al. (1986, Chapter 6))....If you want to get a full essay, order it on our website: Orderessay



If you want to get a full essay, wisit our page: write my essay .