Page 1 of 1

Programming Dynamic Forecasts for Models

Posted: Sun Jul 18, 2010 12:19 am
by Chris_Ger
Hello everybody,

I would like to know, how to program dynamic forecasting (out of sample forecasts) for a model in Eviews 7. My questions in detail:

There is a system of equations or a var object which has been transformed into a model. According to the Eviews manuals you can set the smpl to the estimation period and afterwards set the sample to the forecast period (e.g. forecast-smpl n=2 years -> forecast period is two years) and Eviews will provide dynamic forecasts based the model. These forecasts are shown in a graph.

(1) How can I save the forecasted data in a vector or series in order to use this data for further processing (RSME etc.) in a program?

(2) There is the "forecast" function [eq1.forecast (options) ...] for dynamic forecasts based on equations , but it does not work for models. Is there any comparable function for models which I can apply in Eviews programming?

(3) Given there is no solution to store the data or a comparable function to the "forecast"-function for equations I will have to program the dynamic forecast on my own. My idea is to estimate the model in the estimation smpl, extract the coefficients from the model and calculate the forecasts step by step based on the coefficients. These seems pretty complicated, so maybe someone has a better idea?

Thank you very much and best regards,

Chris

Re: Programming Dynamic Forecasts for Models

Posted: Mon Jul 19, 2010 9:38 am
by EViews Glenn
I'm not quite certain why you think the model won't compute the dynamic forecasts. It certainly does. In point of fact, the model solution options are a superset of those for the forecast.

Re: Programming Dynamic Forecasts for Models

Posted: Mon Jul 19, 2010 9:54 am
by Chris_Ger
Thank you very much, so my understanding is:

1.) model.solve will generate dynamic forecasts (answer to question 2)

2.) the forecast period is set by "smpl"

3.) forecasted parameters are stored in "series"_0 (answer to question 1)

I think I did not correctly understand the "solve" command

Thanks again and best regards

Chris