VAR and transformation of variables

[hk976]

Hello, everyone.
I have a question about VAR. I know each variable should be I(0) and thus i will use log first difference to transform my variables. My question is whether all variables have to transformed in that same manner even though there is a stationary variable without the transformation and another variable that becomes stationary just after putting log. Applying the first difference to all variables uniformly looks right and easy to interpret, but i am afraid i will lose too much information. I will appreciate your consideration. Thank you.

[jmagomez]

Dear hk976,

First, you have to make unit root tests over all variables.

But most important, you have to put all variables that you are working with at a VAR equation without making first difference. And so you can run a Cointegration Test.

You are seeking for similar movements between series. You do not need to make first difference. You are searching also for the same integration order of the variables.

Regards.