Forecasting from a VAR

[Thommo]

How can I forecast from a VAR?

[EViews Gareth]

You can forecast from a VAR by making a model object from the VAR, then solving the model over the period you wish to forecast.

To make a model from your VAR, after you have estimated the VAR click on Proc->Make Model.

To then solve the model, click on the Solve button, and then in the "Solution sample" box, enter the period over which you wish to forecast. Then hit ok.

You will then find some new series in your Workfile with a suffix of _0. These are the forecasted values. Note that by default the model will insert actual values for out-of-sample periods.

[LightMooCow]

Hey, I was wondering if there is also a way to obtain the standard errors or standard deviations also.

[EViews Gareth]

Unfortunately I believe not.

[Gene]

Hey, I was wondering if there is also a way to obtain the standard errors or standard deviations also.

When you solve your model, select the STOCHASTIC option. This will simulation your system. By default, the simulation means of the system will be saved under a new series with a suffix _0m and the standard deviations saved with suffix under _0s.

[super_nova]

In forcasting from Model(VAR)
if my sample is from 1972Q1 to 2007q4
and i want to forcast 2008Q1 to 2013q4 then what should i write into solution sample box?
should i write "2008q1 2013q4"
but it says "error in sample: attempt to set sample outside of workfile range"

[startz]

In forcasting from Model(VAR)
if my sample is from 1972Q1 to 2007q4
and i want to forcast 2008Q1 to 2013q4 then what should i write into solution sample box?
should i write "2008q1 2013q4"
but it says "error in sample: attempt to set sample outside of workfile range"

Before doing the forecast, you need to extend the workfile range all the way to 2013q4 so that EVIews has a place to stick the forecasts.

[arrowstream]

I do have a similar problem..
I have an in sample of 1998q1-2008q4, and want to forecast until i.e 2010q4, i changed the model in proc->structure resize.. to 2010q4.
However, when i run my VAR i just want to include 1998q1-2006q4, and then include 2007q1-2010q4 in the forecast (so that i atleast have 2 years to compare with the original data). So of course i estimated the VAR with 1998-1 - 2006-4, and then did step by step of the above answer, but the problem is that eviews seems to change the data from previous years (i.e comparing the model that eviews solved for and the original one, it differs already in i.e. 2000) which makes the forecast pretty unrealistic..
so the question is how to only make eviews forecast 2007-1 -> 2010-4.

And another problem-in-problem is that when eviews forecast my variables (GDP, inflation and unemployment) i only want to forecast GDP and not the other two variables, so is there some way to include a dummy (or similar) that creates values of inflation and unemployment that follow my past values on these two variables?

i seem to get the best result with the Dynamic-Deterministic Simulation.

[tmlai]

In eviews user interface, there is a "click option" of std dev and bound in solution scenarios & output" when you select the "stochastic" simulation type. However, when I use program code to run the same forecasting, only the parameter mean "_0m" can be generated, no std dev and bound "_0s, _0h, _0l" cannot be generated. Can anyone suggest the relevant example code for model solve with std dev and bound output?

Moreover, I suggest eviews generate the relevant codes used when the users work in each function in user interface. It is very useful for users to program their own codes when some new functions are used. Another statistics program SPSS provides similar functions that make their program more popular.

[trubador]

In eviews user interface, there is a "click option" of std dev and bound in solution scenarios & output" when you select the "stochastic" simulation type. However, when I use program code to run the same forecasting, only the parameter mean "_0m" can be generated, no std dev and bound "_0s, _0h, _0l" cannot be generated. Can anyone suggest the relevant example code for model solve with std dev and bound output?

Try the following:

Code: Select all

modelname.solve(s=a)

[runawayyy]

I estimated a VAR model. Then I clicked Proc - Make model then enter the period over which I want to forecast, and then Solve.

But I wonder how can I view the confidence interval of the forecast. For example, I want to obtain the lower 1% tail value of the forecast density, how can I do this?

I notice that under the Stochastic Option tab in the Model Solution window there is a Interval size (2-sided) box but couldn't figure out what to do next.
Thank you!

[EViews Gareth]

Best thing to do is use the VAR Forecast Add-in. If you want to do is manually though, simply perform a stochastic solve (instead of deterministic), selecting to keep the bounds levels, and, as you say, set the confidence interval level.

[runawayyy]

Thank you Eviews Gareth!
But how can I install the Eviews forecast Add-in ?? I downloaded it but Eviews could not read it!!

[EViews Gareth]

Which version of EViews?

[runawayyy]

I'm talking about Eviews 6 !