Rolling Window Forecast with a VAR Model
Posted: Thu Sep 22, 2016 1:02 pm
Hi,
I’m trying to run a rolling window forecast with a VAR model. I found a program on this forum (at this page: http://forums.eviews.com/viewtopic.php?t=9752) that I took as a starting point, see below.
My situation:
Observations: 212 (sample: 1999m01 – 2016m08)
Fixed estimation period (rolling window): 156 observations (starting with the sample: 1999m01 – 2011m12)
Forecast horizons: 1 to 8
My goal is to obtain 8 series of forecasted observations, each series corresponding to a different forecast horizon. So, the forecast series with h=1 should have 56 observations (from 2012m01 to 2016m08), the forecast series with h=2 should have 55 observations (from 2012m02 to 2016m08) and so on. My question is, starting with the program below, how can I obtain the 8 forecast series?
I’m trying to run a rolling window forecast with a VAR model. I found a program on this forum (at this page: http://forums.eviews.com/viewtopic.php?t=9752) that I took as a starting point, see below.
My situation:
Observations: 212 (sample: 1999m01 – 2016m08)
Fixed estimation period (rolling window): 156 observations (starting with the sample: 1999m01 – 2011m12)
Forecast horizons: 1 to 8
My goal is to obtain 8 series of forecasted observations, each series corresponding to a different forecast horizon. So, the forecast series with h=1 should have 56 observations (from 2012m01 to 2016m08), the forecast series with h=2 should have 55 observations (from 2012m02 to 2016m08) and so on. My question is, starting with the program below, how can I obtain the 8 forecast series?
Code: Select all
'set window size
!window=156
'get size of workfile
!length=@obsrange
'set step size
!step=1
'declare equation for estimation
var var1
'calculate number of rolls
!nrolls=@round((!length-!window)/!step)
'variable keeping track of how many rolls we have done
!j=0
'move sample !step steps at a time
for !i=1 to !length-!window+1-!step step !step
!j=!j+1
' set sample to estimation period
smpl @first+!i-1 @first+!i+!window-2
' estimate equation
var1.ls 1 3 a b
var1.makemodel(mod1)
'out of sample forecast
!ahead=8
'----------------------------------
smpl @first+!i+!window-2 @first+!i+!window-2+!ahead
mod1.solve
for !h=1 to !ahead
if !h+!i+!window<!length then
matrix(!ahead,!length-!window) forecasts_a(!h,!i)=a_0(!h+!i+!window-1,1)
matrix(!ahead,!length-!window) forecasts_b(!h,!i)=b_0(!h+!i+!window-1,1)
else
smpl @first @last
endif
next
next