EViews.com

Posted: **Wed Oct 29, 2014 7:07 am**

Hello all,

Please forgive the simplicity of my question as I am new to Eviews.

I have a data set covering the period 1993q1 to 2006q4. I am trying to generate a series of 1 step ahead ex-post forecasts starting from 2000q1 to 2006q4 using a VAR model. The idea is to estimate a VAR model using the sample 1993q1 to 1999q4 and generate a forecast for 2000q1, then estimate another VAR model with the sample 1993q1 to 2000q1 and generate a forecast for 2000q2, etc., repeating this step until the final date in the data set, then storing the resulting forecasts in a series.

My initial thought was to run the following code to estimate all the VAR models:

Code: Select all

for !i=0 to 24
smpl @first 1999q1+!i
var var01!i.ls lgdp cci

where lgdp and cci are the endogenous variables in the VAR.

However, I get an error message saying that var!i is not a valid name.

Could anyone provide help how I can estimate many VAR models by using a loop command, or suggest a better way on how to tackle this problem?

Thank you.

Posted: **Wed Oct 29, 2014 9:17 am**

You need to add the lags to your VAR.

Code: Select all


create q 1990 2020
series lgdp=nrnd
series cci=nrnd

for !i=0 to 24
	smpl @first 1999q1+!i
	var var01!i.ls 1 2 lgdp cci 
next

Posted: **Thu Oct 30, 2014 1:29 am**

Thank you for your quick reply!

I have successfully estimated the required VARs now, but I am having some trouble generating the forecasts needed and storing them in a series. So far my code looks like this:

Code: Select all

for !i=0 to 24
   smpl @first 1999q1+!i
   var var01!i.ls 1 1 lgdp cci 

   'VAR forecasts 
    var01!i.makemodel(mod01!i)
    mod01!i.solve
    smpl 2000q1+!i 2000q1+!i
    
next

However the resulting series for the forecasts of lgdp, lgdp_0, only includes the forecasted value for 2006q1, but all the other values in lgdp_0 are just the actual observations. It seems that running the loop to solve all the models keeps overwriting the lgdp_0 series for each !i and only includes the last forecasted value in the series. Any idea on how I can store the forecasts for each !i in a series, and not have them being overwritten in the loop?

Thank you!

**Update

I tried modifying the code to the following:

Code: Select all



table fcast

for !i=0 to 24
   smpl @first 1999q1+!i
   var var01!i.ls 1 1 lgdp cci 

   'VAR forecasts 
    var01!i.makemodel(mod01!i)
    mod01!i.solve
    smpl 2000q1+!i 2000q1+!i
    fcast(1+!i, 1)= @elem(lgdp_0, "2000q1+!i")
    
next

with the intention of extracting the forecasts for each !i in the loop, and storing them in the table fcast before being overwritten.

However I get an error message saying that 2000q1+0 is an illegal date.

Posted: **Thu Oct 30, 2014 7:59 am**

I would imagine you want your smpl statement before the solve command.

But your specific problem is that the @elem command doesn't recognise the "2000q1+!i" syntax.

You'll have to do something like:

Code: Select all


%date = @otod(@dtoo("2000q1")+!i)
fcast(1+!i, 1)= @elem(lgdp_0, %date)

EViews.com

Generating series of 1 step ahead forecasts from VAR model

Generating series of 1 step ahead forecasts from VAR model

Re: Generating series of 1 step ahead forecasts from VAR mod

Re: Generating series of 1 step ahead forecasts from VAR mod

Re: Generating series of 1 step ahead forecasts from VAR mod