Mixed orders of integration.

[yanivd]

Is there a way to model variables with different orders if integration using EViews?
In my case it is one variable I(1) and three I(0). Does this necessarily mean they cointegrate?

I can't find (books or otherwise) how one can deal with different orders of integration. I can't trust the cointegration tests because the orders are different.
I also can't say I expect any cointegration because most of my variables are naturally stationary...

Will be happy for any help or clue...

Thanks!

[startz]

I(0) variables cannot be cointegrated because they are not integrated.

[yanivd]

That makes perfect sense :) - silly me.
So, how can one model this situation of one I(1) with three I(0) variables?
Can I just difference the one variable or (over) difference them all?
If I only difference one, how should I interpret the results, say IRFs?