Forecasting with Models

[dhanssen]

I am wondering if the only way to forecast in models is to change the number of observations in the entire workfile in order to create the forecast values for the exogenous variables? This is the only way I know how to do it now, so I just wanted to make sure.

Thanks!

[EViews Gareth]

I'm not sure I follow this question at all.

The EViews model object can be used for forecasting. You forecast with a model by first creating the model, then solving over the observations you wish to forecast. This doesn't, necessarily have anything to do with resizing the workfile.

[EViews Glenn]

I think he means that he wants to forecast outside of the workfile and needs to know whether he needs to expand the workfile and to add values for the observations for exogenous variables.

The answer to those two questions (if indeed I'm interpreting correctly) is yes. All forecasing in EViews (whether from an equation or model) requires that we have someplace in the workfile to put the results, hence the requirement that you expand the workfille. Further, forecasting requires that values for exogenous variables for the forecast period be available. So you'll always have to expand the workfile to include observations that you want to forecast, and make certain that there are valid values for any *exogenous* variables.

Note that lagged endogenous variables, if any, can be generated as part of a dynamic forecast.

[dhanssen]

Yes, that was exactly my question. Great, just wanted to make sure this was the case. Thanks again!

[confused]

Hello,
I am trying to forecast my ARMA model using the out of sample values but I keep getting the error message: 'Unable to compute due to missing data'. Could you please advise me on what must have gone wrong?

[startz]

Hello,
I am trying to forecast my ARMA model using the out of sample values but I keep getting the error message: 'Unable to compute due to missing data'. Could you please advise me on what must have gone wrong?

On possibility is that you are starting the forecast sample more than one after the estimation sample. In this case EViews is missing a lagged value for the forecast.

[confused]

Actually my workfile range is 1964q2-2006q4 and my in-sample range is 1964q2-1984q4. I started the forecast at 1985q1. From what you are saying, I should start it from 1984q4 itself?

[startz]

Actually my workfile range is 1964q2-2006q4 and my in-sample range is 1964q2-1984q4. I started the forecast at 1985q1. From what you are saying, I should start it from 1984q4 itself?

No, what you're doing should be fine.
You might want to post your workfile, including the equation with the problem.

[cnakashima]

Hi,

I'm trying to forecast with a model, but I get an error message "Solve terminated - Solver has stalled: starting values may be too far from a solution". What does that mean?

Thanks!

[EViews Gareth]

Pretty much what it says - EViews wasn't able to find a solution for the model, possibly because the starting values were bad.

[cnakashima]

do you mean the values of the forecast or the values of the original data?
how could I fix it?

thanks!

[EViews Gareth]

Depends on what settings you have set on the Solver page of the Solve dialog.