EViews.com

First, I think it is good to refresh our memory about stationary data by quickly reading this short explanation: http://www.investopedia.com/articles/tr ... z27N8CsGCZ

In case you guys don't want to open the link. Here's the main part: "Using non-stationary time series data in financial models produces unreliable and spurious results and leads to poor understanding and forecasting. The solution to the problem is to transform the time series data so that it becomes stationary"

We can do so by differencing.

I have to be honest and say my knowledge of econometrics is very very limited and that's why I humbly come here to seek propoer technical advice.

Here is the issue. My Dependent variable IS stationary, but my Independent variables are NOT Stationary. When building models/equations on, let's say, eviews, can I keep my dependent variable just with the logarithmic form and my independent variables differencing the logarithms?

Here's JUST AN EXAMPLE of how the equation specification on eviews would look like:

ls logY c dlog_X dlog_Z ar(1)

I hope I can receive some proper advice on this matter as quickly as possible. For some of you this is a walk in the park. I just need to know if leaving the dependent variable without differencing would be ok or not acceptable. In case it is not, what is the way out? (do both sides HAVE to be differenced then?)

Thanks a lot in advance.

Your choice of whether to difference the regressors is based on whether you are interested in knowing whether the level affects the dependent variable vs the change in the level affecting the dependent variable. You might use both.

You should check to see if your regressors are colinear.

What you want to avoid is a "spurious regression" where you have nonstationarity in the dependend and an independent variable.

Would a way to avoid the "spurious regression" in the dependent and independent variable be to perform a causality test (i.e.: Johansen Causality test)? That way I would make sure (if causality has been shown on the result, of course) that a causal relationship exists and a spurious relationships could be ruled out?

Like I said, I am a total newby in econometrics, but from what I have just read, the above would make sense to me. Please correct me if I'm wrong.

I am still trying to learn how to check if my regressors are colinear (collinear, right?).

From what I understand from the first part of your comment, I can use either differenced variables as well as not differenced; but my question remains. Can I use a non-differenced dependent variable (just in its LOG form) and differenced independent variables (Diff_Log)? Or do both sides have to be at level (both sides either differenced or not differenced)?

My dependent variable is stationary, my independents aren't. That's why I wonder if differencing those that aren't could correct the problem of non-stationarity (but that would mean having something like this on the equation specification: ls logY c dlog_X dlog_Z ar(1), where the dependent variable is not differenced and the independents are. Is that duable? Of course the interpretation changes, I am aware of that... but, is it possible to have the dependent variable not differenced and the independents differenced?

Hope to hear from you again soon, and THANKS A LOT for your first post already.