Re: [Gretl-users] Maximum likelihood forecasting
Am 17.01.2019 um 16:54 schrieb Olasehinde Timmy:
Dear Professors,
I will like to know how to compute a dynamic forecast from maximum
likelihood estimation. For example, If I estimated a ad-hoc GARCH model
as follows
series e=inf - beta1-beta2*inf(-1)
series h=var(e)
series h=alpha+theta*e(-1)^n
So how would I forecast dynamically variable "inf" and "h". I would
appreciate if you can share a loop code for this.
Hi,
the part about the levels (first moments) is easy. As long as you don't
have an ARCH-in-mean model or something like that your mean equation is
unaffected. So you have an AR(1) here and you can use gretl's standard
tools ('ols' and 'fcast' basically) to get your forecasts.
About forecasting the conditional variance, I'd suggest to take a look
at the "gig" addon (Garch in gretl) and section 3.5 in its help doc.
good luck,
sven