Simple linear regression question

[Max2004]

Hi there,

I'm currently studying at a foreign university, at which the level of econometrics seems to be very low. As such, we have been given an assignment by our lecturer to do a simple linear OLS time series regression, and he basically just expects us to do it the way he showed on the board (he went through all the basics of OLS, note that this is a Masters course in Finance..).

Anyhow, when I had a look at the data we received, I started wondering: most of the data are in Absolute values and the dependent variable is obviously trended (whether it is a unit root trend or a deterministic one is to be tested, I guess). Also, some (but not all) of the 8 regressors are trended in a similar way. Now, we have been told to just do his very simple version (we're not supposed to think for ourselves, it seems), probably because he doesn't assume any of us know any econometrics, but from my previous econometrics courses, I recall that it would be crucial to first make both the dependent variable and the regressors Stationary, before attempting a regression.

As such, I figured the right approach would be to separately test all the variables for unit roots, e.g. by ADF - and if one was found, use the first-difference of the variable instead of the absolute value in the regression. Also, if e.g. one of the variables turn out to suggest a deterministic trend rather than a unit root, I assume it would be fine to regress the variable on a trend and then use the residuals from that regression (Frisch Waugh).

Now, usually I would've just done the above, but since the lecturer has given us a potentially fundamentally wrong assignment (to 300 students), I wanted to be sure before I went ahead and changed the variables etc. I hope some of you can tell me if I'm right or wrong :)

Best regards,

Matt

[mectricsdonk]

he prolly wants to prove you guys a point or something... but if the problem isn't cleared up by next lecture, find a new uni