EViews.com

I am interested if there is an option to do mixture of forecasts in eviews:

Lets say you have two ways of forecast: f1, f2 and we want to combine them. Instead of making a regression from the whole sample of each one, can we perform sub-regressions to sub-samples and combine the best?

It would be something like a selection-based model but with some criteria, like sic,aic? Can this be done in eviews ?

You can, of course, estimate two different regressions over two different samples, and then forecast each of those regressions over two different samples. Then it is relatively simple to stitch those two forecasts into one series.

Thank you for that! I will try it!

Is there any model to implement selection criteria; To establish which sub-regression is the best? Or I have to seek that manually from the results on the each regression; should I check the R^2 or some of the f,t -statistics? This more of a 'quality' question I guess!

Thanx a lot!