fGarch: possible bug(s) and suggestions
Yohan Chalabi
chalabi at phys.ethz.ch
Wed Feb 20 16:26:55 CET 2008
>>>> "MM" == michal miklovic <mmiklovic at yahoo.com>
>>>> on Wed, 20 Feb 2008 06:27:46 -0800 (PST)
MM> h = c(rev(spec at presample[, 2]), rep(NA, times = n))
MM>
yes, this is pretty obvious, thanks to point it out.
MM> That is, the second column in presample should not be raised
MM> to the power of delta because it is conditional variance and
MM> not conditional sigma as I inferred from the garchSpec function.
MM>
MM>
MM> Second, I would suggest to slightly alter the definition of
MM> 'getting and extending series' in the predict function in the
MM> following way:
MM>
MM> # Get and Extend Series:
MM> x = c(object at data, rep(mu, M))
MM> h = c(object at h.t, rep(omega, M))
MM> z = c(object at residuals, rep(0, M))
I think it is a matter of notation.
It is common to drop the constant terms in the mean and variance
equations (mu, omega). So, the code makes sense.
regards,
Yohan
More information about the Rmetrics-core
mailing list