I want to estimate a var model the variables are stationary in I (1), when I do the estimation and I look for the optimal lag I find that the test lag length creteria the AIC criterion chosen "0", when I designate the number of delay to three he chooses 3 when I design 4 he chooses lag 4 and when I design 5 he chooses lag 5 when I choose 6 he writes log of not positive number, knowing that when I choose lag 3 or lag4 or lag5 the model becomes no stationary, I don't know what decision I have to make, these figures show the estimated tests with lag 1 and lag 3
At first you must run the co-integration test to examine the long run relationship among the variables, and if the variables are co-integrated you can not apply the VAR model and you must run the VECM model or the ECM equation as you need from your work.
for the VAR model you estimte did you use the first difference series or the level series??? because you cannot use the original series in VAR model since they are I(1) series.
I used the first difference, and i estimate johanson cointegration there is relation at the long run, but when i used VECM with 1 lag i got this results
so you can not use the VAR model for the short run term you must use the VECM model
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Writing