My model is a blue, it's a simple OLS regression like Y: X Z W

X Z W are standard data (latest aviable) but Y variable uses realtime data.

So there's a way i can run fcast --rolling command with dependent variable changing at every new step.

Thanks

Francesco 

2015-09-22 8:37 GMT+02:00 Sven Schreiber <svetosch@gmx.net>:
Am 22.09.2015 um 01:27 schrieb Francesco Raffaele P.:
> Hi,
> I need to run fcast command on Vintage Data.
>
>
> I explain myself.
> For the dependent variable, every month there is a new observation and
> the previous last is changed by revision.
> So i need to run a recursive prevision using fcast command, but every
> time gretl make an estimate ex 01:1991-02:1991-03:1991 the dataset is
> changing.
>
> Where i can find the fcast command code? Maybe i can try to change it.
> Is there some extension for this?
> Someone can help me to do this?

'fcast' of course cannot "automagically" produce forecasts out of the
blue. You need to specify a model first. So you should first ask
yourself what is the appropriate model for this kind of data. This is
something only you can determine, not gretl, because tastes and
assumptions differ between researchers.

Having said that, perhaps you find chapter 8 of the user guide on
"realtime data" helpful. This might help you to update your changing
dataset at every step.

Once you have your dataset updated (in a loop, presumably) and your
model re-estimated, 'fcast' should work in the usual way.

hth,
sven
_______________________________________________
Gretl-users mailing list
Gretl-users@lists.wfu.edu
http://lists.wfu.edu/mailman/listinfo/gretl-users