Re: [Gretl-devel] OLS Forecasting
On Fri, 6 Feb 2009, Henrique wrote:
> On Thu, 5 Feb 2009, Henrique wrote (in user-list):
>
> I had estimated a OLS using first difference variables, but I would
> like to forecast the dependent variable in its level. I know that this is
> possible on EViews, where we can estimate a regression D(y) = a + b D(x)
> (where "D" states for the difference operator, "a" and
"b" are parameters,
> and "y" and "x" are the variables) and obtain the predicted
"y" values
> automatically. How can I do this on gretl?
And Allin had answered me:
> On Thu, Feb 5, 2009 at 7:06 PM, Allin Cottrell <cottrell(a)wfu.edu> wrote:
>
> Gretl will do that automatically for an ARIMA model, but not
> otherwise. But you have the cum() function to integrate a
> differenced series.
I would like to have this functionality for OLS and VAR. What do you
think? Is it possible?
This is now implemented on a trial basis in CVS/snapshot, but only
for single-equation OLS at this point. If you estimate a model
where the dependent variable is recognized as the first difference
of another variable, you get the option to integrate (i.e.
produce a forecast in levels).
This is available in the GUI forecast dialog, and also via the
command-line flag --integrate for the "fcast" command.
Testing of this would be nice!
Allin.