Hi Matthieu,
I am interested in using your package to test for cointegration in VAR models. One advantage of your package seems to be that unlike urca::ca.jo(), tsDyn::rank.test() allows for VECM models that have only one lag in levels / no lagged differences in the VECM representation. However, this prints out a warning that many downstream methods can't be trusted, even predict(). I was wondering whethere there has been any progress or whether this is still unreliable.
Thanks a lot for any help and for the very useful package!
Hi Matthieu,
I am interested in using your package to test for cointegration in VAR models. One advantage of your package seems to be that unlike
urca::ca.jo(),tsDyn::rank.test()allows for VECM models that have only one lag in levels / no lagged differences in the VECM representation. However, this prints out a warning that many downstream methods can't be trusted, evenpredict(). I was wondering whethere there has been any progress or whether this is still unreliable.Thanks a lot for any help and for the very useful package!