Re: [Gretl-users] CARR model MLE
On Thu, 23 Feb 2012, Dan Běsoň wrote:
I forgot - is there a way to extract the in-sample
predictions of the CARR model?
Please don't dump the gretl-users digest into mail to the
list.
Predictions for which variable? Once you have your parameter
estimates for the CARR model you can calculate any relevant
series you like. E.g.
series lambda = mean(Rng)
series lambda = a + b * Rng(-1) + c * lambda(-1)
series Rng_norm = Rng/lambda
(This assumes that a, b, and c are the values optimized by
mle; the optimal values are also accessible as the $coeff
vector.)
There are no "fitted values" as such for a model estimated via
mle, since you're specifying a likelihood function, not a
dependent variable.
Allin Cottrell