unbalanced panel data and relevant variables

[Maddi]

Hi @ all,

first of all, sorry for my English. I am a German girl.
I don't have any experiences in programming with Eviews. But unfortunatelly, I have the following problem now:
I have unbalanced panel data with different variables. I would like to select, which variables are really important and deliver the highest degree of information (BIC, AIC). Eviews should first present the variable with the highest information, then select the second one, etc. till no further variable explains better the regression. I don't have a clue how to do this in Eviews. I really hope, someone can help me.

Thanks a lot in advance!!!

[EViews Gareth]

EViews has nothing built in that will let you use AIC or BIC as the selection criteria. However you could use stepwise least squares, which lets you select variables based on R^2 or t-statistics.

[Maddi]

Thanks a lot!!!! This is really good to know!! Your suggestion would mean, I have to try different variables and to select the best combination in respect to t-stastic or R^2?

[EViews Gareth]

EViews will do the selecting for you.

[Maddi]

Sorry for annoying you! Unfortunatelly, I am still not sure how to do this. Can I do this just by clicking through the menu bar or do I have to write a code?
Up till now, I have done some unbalanced panel regression by selecting some variables by my own. Then used the menu bar to estimate my regressions. But I chose the combination of variables. This doesn't make really sense, because there a such a lot of possible combinations, so it would be great, if there was another opportunity.

Thank you very much!! I know, I am really annoyiny. Sorry.

[EViews Gareth]

From the menus. Choose STEPLS as the estimation method.

[Maddi]

Thanks! I have tried - unsuccessful. Can it be that StepLS is only applicable for multiple regression and not for pooled data?

[EViews Gareth]

You said you were using panel data, not pool data.

[Maddi]

Sorry, statistic is too complicated for me. I am really a stupid girl in respect of this.
I read papers about panel data and then I found a description how to estimate this in eviews. But it was a pool estimation. Now, I changed my pool data to panel data and how surprisingly..your suggestion works!!! :-) Thanks!
But to be honest, I am still not really sure what are the big differences between panel and pool data. If I use the OLS, the results will be the same.