Dynamic model selection

[RalphVerhoeks]

Dear Eviews expert,

I have a question about the program attached. This program is designed to apply a rolling regression. The program selects each point in time the best model, according AIC, BIC, Adj R2 and PCS, which is used to make 1-step ahead forecasts.
As an additional model selection criterion I would like to do the following: estimate the full model with all possible regressors each point in time and retain only the 3 variables in the model with the highest t-statistics. So, only I want to hold the variables in the model which rank the highest in t-statistics. Then, I would like to use this model to make 1-step ahead forecasts. How can I program this dynamic model selection?

I hope you can help me and thank you very much in advance!

[RalphVerhoeks]

Can anyone of Eviews help me with my question?

[EViews Gareth]

Have you thought about using stepls to do the t-stat based model selection?

[RalphVerhoeks]

dear eviews gareth,

I know the technique of stepwise regression. However, I only want to include the 3 highest t-statistic variables in my model, each point in time, to make in more flexible.
Can you me how to program that in the program attached in my earlier post?

Thank you very much in advance!

[EViews Gareth]

stepls lets you specify 3 regressors with the three highest t-stats...