Hi,
I want to fit standardized generalized hyperbolic distribution to my data. I am
aware, that I can do this with the dsgh command of the fBasics package along
with the optim command. My problem is, that I also want to have a derivation of
it. So I need the theory behind it, i.e. I need the formula of the probability
density function which they use and the derivation of it.
I thought about standardizing the generalized hyperbolic distribution. So I use
the formula of the mean and the variance (e.g. can be found on wikipedia) and
set them to zero and one. Then I try to solve for single paramters and insert
them in the original pdf and use this along with the optim command. BUT the
problem is, that there is no unique solution, so the mean and the variance, as
you can see on the wikipedia page depends on several parameters and terms. So I
could have different set of parameter combinations which all fulfill the
requirement of the mean to be zero and the variance to be equal to one. What
are people doing in this case? How do they get a "unique" solution for the
standardized version?
e.g. I do a simplified example, so you see, what I mean
Suppose, the pdf is given by
mu + alpha + 2* beta + 3* delta
the mean is given by
mu + delta*beta
and the variance given by
beta*delta + delta/(alpha-beta)
I set them to zero and one:
0 = mu + delta*beta
1 = beta*delta + delta/(alpha-beta)
now I can solve in different ways and insert in different ways in the original
pdf.
So the result is, that I can get a pdf formula, which depends on alpha, beta,
and delta. The mu is fixed.
Or I can get a pdf, which depends on mu and delta. The alpha and delta is then
fixed.
Both would fulfill the requirement of mean zero and variance one.
What should one do in such a case?
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
