VAR and Granger Causality

[lnp3]

1) what are the stepwise procedures for granger causality in time-series data? First, I check the stationary and co-integration properties of variables. If all variables are first-difference stationary and co-integrated (after Johansen test), then I go for granger causality

2) Where does unrestricted VAR fall in this picture? 3) Where does VECM fall in this picture? I know that we use VAR when first difference stationary variables are NOT co-integrated. VECM is a form of "restricted VAR", meaning that we use it when variables are co-integrated.

Which one should I use for Granger causality? What is the purpose of unrestricted VAR? I truly appreciate your help on this!

best regards,
MD
:(