Model Solve - smooth transition from historical to forecast?

For technical support, tips and tricks, suggestions, or any other information regarding the EViews model object.

Moderators: EViews Gareth, EViews Moderator

gg_pse
Posts: 5
Joined: Tue Feb 07, 2012 3:12 pm

Model Solve - smooth transition from historical to forecast?

Postby gg_pse » Tue Feb 07, 2012 3:32 pm

I am fairly new to eviews and would appreciate any help on this topic. I have a system of equations in the MODEL object that I am trying to solve simultaneously and create forecasts of. The problem is that for the starting value of the forecast of a given endogenous variable there is an anomalous dip which throws off the entire series and subsequently all the other variables that are dependent on it. For example a variable showing population growth which has historical data till 2011, suddenly turns negative in 2012. I am trying to figure out how this first forecasted value for an endogenous variable is created, and how I can modify it? Is there a way to use add factors to get a smooth transition from historical data to forecast?

The simulation type I am using is deterministic (dynamic solution) and the solver is set to Preferred solution starting values - "Actuals" and Forward Solution "User supplied in Actuals".


I would greatly appreciate any help. Thanks.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12298
Joined: Tue Sep 16, 2008 5:38 pm

Re: Model Solve - smooth transition from historical to forec

Postby EViews Gareth » Tue Feb 07, 2012 4:45 pm

Hard to say without knowing more about the specification of the model/equations
Follow us on Twitter @IHSEViews

gg_pse
Posts: 5
Joined: Tue Feb 07, 2012 3:12 pm

Re: Model Solve - smooth transition from historical to forec

Postby gg_pse » Wed Feb 08, 2012 9:53 am

I am attaching 2 txt files here. The first has all the text view of the entire model. The second has the equations defined which do not display in the text view in the first file.
Thanks
Attachments
Equations.txt
(6.16 KiB) Downloaded 309 times
Model.txt
(1.85 KiB) Downloaded 299 times

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12298
Joined: Tue Sep 16, 2008 5:38 pm

Re: Model Solve - smooth transition from historical to forec

Postby EViews Gareth » Wed Feb 08, 2012 9:56 am

Could you post the workfile?
Follow us on Twitter @IHSEViews

gg_pse
Posts: 5
Joined: Tue Feb 07, 2012 3:12 pm

Re: Model Solve - smooth transition from historical to forec

Postby gg_pse » Wed Feb 08, 2012 10:21 am

Unfortunately I can't post the workfile because it has forecast data from a third party that is proprietary. I was wondering if Icould get some ideas on how to proceed, based on the equations and Solve properties of my model that I have posted.

Currently if I set the scenario to "Baseline" it shows bad values in the first period of forecasts leadig to a sharp dip in every forecasted variable. If I change the settings to "Actual" and check "Exclude for periods where Actuals exist", then I get the forecast to be a straight line where the last historical value is repeated throughout the entire forecast period.

EViews Gareth
Fe ddaethom, fe welon, fe amcangyfrifon
Posts: 12298
Joined: Tue Sep 16, 2008 5:38 pm

Re: Model Solve - smooth transition from historical to forec

Postby EViews Gareth » Wed Feb 08, 2012 10:39 am

What happens if you do a solo forecast from the equation, rather than using the model object?
Follow us on Twitter @IHSEViews

gg_pse
Posts: 5
Joined: Tue Feb 07, 2012 3:12 pm

Re: Model Solve - smooth transition from historical to forec

Postby gg_pse » Wed Feb 08, 2012 10:45 am

In that case it creates forecasted values over the historical period, but cannot forecast out in the future since the independent variables also do not have values in the future periods.

EViews Chris
EViews Developer
Posts: 161
Joined: Wed Sep 17, 2008 10:39 am

Re: Model Solve - smooth transition from historical to forec

Postby EViews Chris » Wed Feb 08, 2012 11:21 am

It's very hard to give any specific advice.

If your model is producing forecasts that you don't like, you can either:
1) change the equations of the model
2) change the forecasted future paths of any exogenous variables
3) include 'add factors' in the model (really just exogenous variables with a special purpose) that reflect tweaks to the model that reflect your judgement as to why the model is not adequately matching up to reality

If you think a particular equation seems to be the cause of the problems, you could verify that by dropping the equation from the model and temporarily treating the endogenous variable of the equation as an exogenous variable and then entering a future path for that variable by hand. If that makes everything in the forecast look sensible, then you've narrowed down the problem to the equation that has been dropped and you can try to investigate further why it is forecasting poorly. That might give you somewhere to start.

gg_pse
Posts: 5
Joined: Tue Feb 07, 2012 3:12 pm

Re: Model Solve - smooth transition from historical to forec

Postby gg_pse » Wed Feb 08, 2012 11:35 am

thank you, i will give these a try


Return to “Models”

Who is online

Users browsing this forum: No registered users and 1 guest