Hi, I carried out an analysis using the VAR (VECM) framework. My variables are integrated of order 1 and are cointegrated (based on the result of the trace statistic and maximum eigenvalue), However, I noticed that the VECM is not stable (as some points lie outside the unit circle). How do I correct this please?
Well, the non-stationary model associated to the VECM should have unstable roots, BUT its companion stationary representation should have stable roots.If you say that this companion model has unstable roots, then you made some invalid assumptions about the stationarity of at least one of variable's transformations (say its first difference or its second difference). In other words, you have taken at least one I(2) variable as a I(1) variable OR at least one I(1) variable as a I(0) variable. Once you make the corrections, the companion stationary representation should have stable roots.
Thank you so much Carlos. I'll take a look at the unit root test results again.
Such issue arises when we misunderstand the actual order of integration of original time series stochastics process. you should one again examine the order of integration of you actual data
- Badges
- Science topic
- Similar topics
- Correction