Stepwise regression

[ami]

Hi all,

I am quiet new on Eviews and would appreciate some guidance.
I want to forecast GDP growth. To do so, I selected a certain number of variables, and plan to regress the GDP on them. I am supposed to use several equations, depending on the month of the year. It seems that the most appropriate method to select the more pertinent variables is to use stepwise backwards. But I have no idea of how to program this command.
How do I know which variables always include and which one are the potential?
Should I put all the variables at the same time, or process by step?

I am quiet lost, so would be very grateful if someone could shed the light on this topic..

Thanks in advance

[EViews Esther]

Stepwise least squares regression is a built-in function in EViews (Page 48 of EViews Uses Guide II).

[ami]

Hi Esther,

Thanks.
I already read the stepwise's section in the Help Guide.
The problem is that I have created my workfile, with roughly 130 observations. But now I want to know which variables are the most relevant. I want to test 12 equations (1 for each month of the year) to regress the GDP on the best variables, in order to have the best GDP growth forecast possible.

So should I creat a group XS for my variables, just put Y (the GDP) C in the first dialog box, and put XS in the other one?
Contrary to the exemple provided in the help guide, I do not known a priori which variables to use for a linear combination of Y.

Whats is the bets way to proceed?

Thx

[EViews Esther]

So should I creat a group XS for my variables, just put Y (the GDP) C in the first dialog box, and put XS in the other one?

yes.

Contrary to the exemple provided in the help guide, I do not known a priori which variables to use for a linear combination of Y.

I am a bit confused. Don't you have a list of variables to be used as the set of potential regressors?
If you want to get a broad idea of selecting the number of regressors, you can run the Bayesian Model Averaging addin (named BMA, downloadable at http://eviews.com/Addins/addins.shtml). Please note this is not a model/variable selection method, however, you will get an idea which variable could be included in your model. I *only* recommend you to read the example section of the documentation.