How will the model stability problem be addressed? Are these tests necessary to diagnose the model stability in the case of (ARDL)?
Veering of the cusum or Cusum square of the grid is an indication of model instability. It is imperative to conduct model specification instability test in time series as most time series macroeconomic data are usually perturbed with structural breaks at some point in time. The situation in your case might be a case of structural break in your series. You can plot the graph of your dependent variable to check for a discernable break in the trend or use the chow test to test the probable break dates.. After that try to model the break date as a fixed regressor in eviews. It should solve the problem....
Yes, the model is not structurally stable so you can also perform chow test to check for presence of a structural break.
According to cusum and cusum square the model is unstable. Please check the stationarity of the variables once. Chow test can be done once the break points are predetermined.
be careful when doing stationarity tests, if you have structural breakpoints in your data the results of conventional tests (DFA, KPSS, PP...) may be misleading. You have to use unit-root tests with structural breaks. And of course, your graphs indicate that your model is unstable.
@Chiraz Karamti, you are right with the weakness of the conventional unit root and stationarity test during structural breaks. But I believe the break dates suggested by these unit root test might not actually be the correct period of break in the affected series.. I usually suggest a chow test, graphical observation of a discerning shift or an history of a financial or economic phenomenon (e.g economic or financial crisis)
yes absolutely, but the unit-root tests with breakpoints give graphics (in STATA) with the dates of breaks so you can judge if the test considers the appropriate breaks or not. For the chow test, it has to sequential test.
Dear Zeravan,
The model is actually unstable.
Time series data are prone to structural breaks so I will employ you to use chow test for verification.
Also conduct other tests as mentioned above.
Regards
Yes your model seems to be unstable as the CUSUM square line is outside the boundary. we face such situation many time where CUSUM line is ok but CUSUM square is outside. I think that there could be structural break in your data which you might have not checked, so i would suggest for structural break test, which can be done through multiple test process such dummy variable, chow test and if you want to get advance level you can apply Andrew Zivot.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs