Closed Paviaym closed 3 years ago
Hi Brian
Disclaimer: it has been a while I haven't used this model, and I am not up to date with the literature
I am not aware of specific criterion for lag length selection in the threshold unit root context. I would rely on tools for linear unit root tests. I think a major contribution was Ng and Perron (2001) modified AIC (not available in tsDyn
), but there is surely something new in the meanwhile. See also this Stack post
Dear Matthieu,
Thank you for your response. I will check the post and paper you shared.
Best regards
Paviaym
On Mon, Dec 7, 2020, 23:31 Matthieu notifications@github.com wrote:
Hi Brian
Disclaimer: it has been a while I haven't used this model, and I am not up to date with the literature
I am not aware of specific criterion for lag length selection in the threshold unit root context. I would rely on tools for linear unit root tests. I think a major contribution was Ng and Perron (2001) modified AIC (not available in tsDyn), but there is surely something new in the meanwhile. See also this Stack post https://stats.stackexchange.com/questions/147956/lag-selection-for-augmented-dickey-fuller-test
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/MatthieuStigler/tsDyn/issues/30#issuecomment-740163008, or unsubscribe https://github.com/notifications/unsubscribe-auth/ASATPJYSOUYSFRD44SVL353STU3SHANCNFSM4UPM3PVA .
I could not see a function about the determination of lag length for unit root tests, for instance for BBC or kapshintest. Can you please recommend a way to decide on the lag length in the execution of these tests? I was happy to see lags.select but later noticed that it is for cointegrated vectors.