I need to make a VAR model with variables: industrial production index, cumulative sanctions index, exchange rate, oil and gas prices. The ADF test shows stationarity for the type without drift and trend for the first differences of variables, however, the results of the VAR model give large information criteria and the absence of autocorrelation only with large degrees of freedom when using these variables. Should I move on to the specifications of the VAR model or do I need to further work out the variables (make adjustments)? Is there any checklist that I can follow in order not to get into a mess
Hi Alice, what do you mean by "absence of autocorrelation only with large degrees of freedom"? It is not clear to me why you are looking for some absence of autocorrelation in the variables, do you mean in the model residuals perhaps?
Btw, the easiest and fast way for choosing the VAR order is (as you probably already know) through information criteria (IC). I am not aware of the existence of some "threshold" IC values suggesting that your selection could unreasonable. You can estimate parameters properly with OLS as long as your variables are stationary, then the tricky part is about how to conduct structural analysis properly. Any suggestion about how to move forward really depends on the specific kind of analysis you are going to do.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning