Implementing unit roost tests in ARMA models of unknow order
This paper compares the performance of classical and recent unit root tests based on different estimation procedures, including fitting ARMA models of unknown orders. The article also introduces an estimator of the spectral density function that is based on the estimation of an ARMA model with data previously detrended by GLS. The Monte Carlo experiment shows that tests improve their performance if an ARMA model is estimated, instead of an autoregressive approximation. The best results are obtained by tests based on the estimation of the spectral density function.
KeywordsOrdinary Little Square Unit Root Test ARMA Model Error Correction Model Generalize Little Square
Unable to display preview. Download preview PDF.
- 18.Sánchez, I. (2002), Efficient Tests for Unit Roots and Prediction Errors,J. Statis. Plan. and Infer. Forthcoming.Google Scholar
- 21.Stock. J.H. (1990), A Class of Tests for Integration and Cointegration, Manuscript. Harvard University.Google Scholar