Classic tests for stationarity of time series and of cointegration may fail when data are observed with dominant noise. We show that results of standard ADF are biased towards more stationarity, while the Johansen’s test cointegration test may produce unreliable results and generate false long-run signals. We show that data filtering improves the performance of standard tests and so it should become a good practice when dealing with very noisy datasets. We prove the effectiveness of different filtering strategies using simulated series.
Trends and long-run relations in cointegrated time series observed with noise / Gianfreda, Angelica; Maranzano, Paolo; Parisio, Lucia; Pelagatti, Matteo. - (2020). (Intervento presentato al convegno SIS tenutosi a Pisa nel 2020).
Trends and long-run relations in cointegrated time series observed with noise
Angelica Gianfreda;
2020
Abstract
Classic tests for stationarity of time series and of cointegration may fail when data are observed with dominant noise. We show that results of standard ADF are biased towards more stationarity, while the Johansen’s test cointegration test may produce unreliable results and generate false long-run signals. We show that data filtering improves the performance of standard tests and so it should become a good practice when dealing with very noisy datasets. We prove the effectiveness of different filtering strategies using simulated series.
Pubblicazioni consigliate
I metadati presenti in IRIS UNIMORE sono rilasciati con licenza Creative Commons CC0 1.0 Universal, mentre i file delle pubblicazioni sono rilasciati con licenza Attribuzione 4.0 Internazionale (CC BY 4.0), salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris
