- Check stationarity of time series using ADF or PP test.
- Check co-integration using Johansen co-integration test.
- If you find any co-integration equation, go for VEC model. However, VAR is fine too for simplicity. Estimation parameters do not differ drasticially.
- Decide on lag orders.
- Perform diagnostic check.
- If all good, perform out-sample forecast.
- Check accuracy of both in-sample and out-sample forecast.
Video on how to perform VAR and VEC models in R are available on ResearchHUB YouTube channel.
See published studies using VAR/VEC models: