[R-SIG-Finance] Copula Package

[salmajj]

Christophe Dutang <dutangc <at> gmail.com> writes:

> 
> Hello,
> 
> I think what you need to look at is the log-likelihood. Something like sum
(log(dcopula(gumbelCopula(3),
> x)). 
> 
> By the way, there is also the gumbel package available on CRAN where some 
classic fitting methods are available.
> 
> Regards
> 
> Christophe
> 
> Le 9 nov. 2010 à 22:20, salmajj <at> softhome.net a écrit :
> 
> > Hi I tried a lot of time to send this message hope it works this time!
> > Hi everybody,
> > my objective is to find the corresponding parameter of the gumbel copula 
that best fit my empirical
> dependancy structure. i.e I already know that the gumbel copula is the best 
family but i want to decide on
> the best parameter ? 
> > in other words
> > let
> > x <- rcopula(gumbelCopula(3), 100) 
> > suppose we do not know that alpha=3 is the right value and we are 
wondering if the gumbel copula with alpha
> equal to 2 is a good fit
> > let test :
> > gofEVCopula(gumbelCopula(2), x)
> > this returns
> > Parameter estimate(s): 3.044912
> > Cramer-von Mises statistic: 0.0004143588 with p-value 0.8616384 
> > let test :
> > gofEVCopula(gumbelCopula(3), x)
> > this returns:
> > arameter estimate(s): 3.044912
> > Cramer-von Mises statistic: 0.0004143588 with p-value 0.8556444 
> > So if I well understand this function gofCopula only indicate that the 
gumbel family is a good fit and we can
> not decide on the best parameter of the gumbel copula as the results were 
the same!
> > So which test we could use to know that actually the gumbal copula with 
parameter 3 and not 2 fit best the dependancy?
> > Thanks a lot!
> > 
> > _______________________________________________
> > R-SIG-Finance <at> stat.math.ethz.ch mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-sig-finance
> > -- Subscriber-posting only. If you want to post, subscribe first.
> > -- Also note that this is not the r-help list where general R questions 
should go.
> 
> --
> Christophe Dutang
> Ph.D. student at ISFA, Lyon, France
> website: http://dutangc.free.fr
> 
> 


Hi all, 
I understand that rmvdc generates random number from mvdc object. But the mvdc 
object can only be used if we define the marginals! So my question is suppose 
we don't find any distribution which fit marginals so we use the Canonical 
Maximum Likelihood method (This approach uses the empirical CDF of each 
marginal distribution to transform the observations into pseudo observations 
with uniform margins) SO after finding the copula which fit the dependancy HOW 
i can generate random number which mimic the data? 
Hope my question is clear, please if someone have an idea help me!