No, first fit the standardized residuals with the distribution you want (e.g. the spd), then apply the cdf transform on those residuals(i.e. pspd(std.resids, fit) ) to get the pseudo-uniform numbers, then collect into a matrix and fit using the copula. Once you have fitted, use the random number generator of copula to get a correlated sample which you then transform back by using the quantile function (qspd), and then reintroduce them into the univariate garch fit from stage 1 from which you can now simulate. HTH. Alexios
Yana Roth wrote:
Yes, I filtered the residuals with GARCH. You mean, I should apply
fit.copula.rank to standardised residuals directly?
I thought, I should estimate CDFs before....
Thank you
--- On *Fri, 1/30/09, alexios /<alexios at 4dscape.com>/* wrote:
From: alexios <alexios at 4dscape.com>
Subject: Re: [R-SIG-Finance] Copula in R
To: yana.roth at yahoo.com
Cc: r-sig-finance at stat.math.ethz.ch
Date: Friday, January 30, 2009, 3:05 PM
The QRMlib function "fit.tcopula.rank" with method="kendall"
will accept
the uniform data from the cdf transformation.
It is more commonplace to first filter the data with a process like
garch, and then apply the fit to the standardized residuals.
There is also a package on r-forge for the semi-parametric distribution
with pareto tail which implements density, distribution, quantile and
sampling (http://r-forge.r-project.org/projects/spd/).
-Alexios
Yana Roth wrote:
> Hello,
> I try to reproduce copula fitting from Matlab by R. I constructed pieswise
distribution: Generalised Pareto at the tails and empirical distribution
estimated with Gaussian Kernel. Like this I obtain 15 CDF. However, I dont find
my way to convert them to uniforms and fit copula.
>
> If you could provide some help, I would be thanjful
>
> Thank You
> Yana
>
>
>
> [[alternative HTML version deleted]]
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> 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.