Dynamic forecast

[Chris]

Hello,

I have a simple eq of the form: Dpt= c(1)+c(2)*Q(t)+c(3)*(Dpt(-1)+Dpt(-2))/2 and I desire to employ from a range of 314 observations an out-of sample (270-314 obs) dynamic forecast with 6 steps ahead and compare it with a random walk static one. However, I saw that it is not possible to specify the number of steps in the forecast window. Even though similar topics exist, I couldn’t identify the right procedure. In the command reference, at dynamic forecast from an eq section there is no option regarding step numbers, in contrast with Sspace forecast section. Does it require then a state space model construction along with Kalman filters or how can I deal with this issue?

Your help will be highly appreciated!

[Chris]

I’ve managed to put together a code but I’m not sure if the outcome is a dynamic 6 step ahead out of sample forecast. It looks more like 1 step forecast with a rolling window of 6. First, eq1 was estimated for in-sample period (until 2005:02) and afterwards:

' Sample forecast
smpl 2005:02 2008:10
'get size
!length = @obssmpl
'define series
series yhat
series yhat_se
' move sample 6 steps ahead (6month forecast)
for !i = 1 to !length step 6
smpl 2005:02+!i 2005:02+6+!i
' make forecasts
eq1.forecast(f=na) tmp_yhat tmp_se
' copy data in current forecast sample
yhat = tmp_yhat
yhat_se = tmp_se
next

Can somebody provide me some indications as I really need to come up with these results.
Thanks!