rolling forecasting

[hichbenn]

Hello dear friends,

I wish I can get a response to my query:

I would like to estimate a VAR model from 1997q1 to 2010q1, then calculate the RMSE from 2007q1 to 2010q1 for three horizons: 1, 2 and 3 quarters ahead.

Eg. I have to estimate the VAR from 1997q1 to 2006q4, then calculate the forecasting from 2007q1 to 2007q3, then restimate the var again from 1997q1 to 2007q1 then provide forecasts from 2007q2 to 2007q4,.... until the end of the sample. Finaly, I will calculate the RMSE to evaluate the model VAR chosen.

This is feasable in eviwes manually but demand alot of time, so I would like to have any help to make a program in eviews which can solve it.

Thank you.

[EViews Gareth]

Take a look at the rolling estimation/forecasting programs and examples. Things are a little more tricky in a VAR since there is no forecast procedure built into the VAR, so you'll have to make a model out of your VAR at each roll, then solve the VAR over the period you want to forecast.

[hichbenn]

Can you please send me a link for this topic.

I didn't get your idea, please would you tell me more details?

I mean I need to have the whole VAR if I have to calculate forecasting for any variable.

Thx alot

[EViews Gareth]

The first four posts in this very forum are all rolling regression related. But essentially, you want this post:
http://forums.eviews.com/viewtopic.php?f=15&t=878

Once you've mastered, and understood how a rolling regression program works, it is trivial to change the part that does the estimation at each step into something that estimates a VAR at each step rather than a regression at each step.

[hichbenn]

thank you my friend,
I got you but I think I will have the program myself ;) as there is nothing ready for now :)
thx alot