This paper proposes a new portmanteau-type statistic by combining several lags of the sample autocorrelations to test for the presence of a unit-root of an autoregressive model. The proposed method… Click to show full abstract
This paper proposes a new portmanteau-type statistic by combining several lags of the sample autocorrelations to test for the presence of a unit-root of an autoregressive model. The proposed method is nonparametric in nature, which is model free and easy to implement. It avoids modeling the fitted residuals and does not require estimation of nuisance parameters, as commonly done in the augmented Dickey–Fuller or Phillips–Perron procedure. Asymptotic properties of the test are established under general stationary conditions on the noises. Finite sample studies are also reported to illustrate the superior power of the proposed method. Applications to test for cointegration are also given.
               
Click one of the above tabs to view related content.