[R-SIG-Finance] rugarch and copulas

Ludovic Theate ludovic.theate at gmail.com
Wed Oct 24 13:14:07 CEST 2012


Hi everybody,

I recently discovered the rugarch package and I have some questions
about how I could use it for the model I am interested in.

In fact, we develop a currency model, using in a first step, some
garch models in order to fit a basket of currencies change rates. Then
we try to account the correlation between each of them using a copula.
A first version of this model has been developped using the fGarch
package.

Question 1 : If I have correctly understood the use of the fGarch
package, I have to fit the model to the data using the garchFit
method. Then in order to simulate some trajectories, I need to use the
garchSim method which needs a model specification created by the
garchSpec method (which requires to extract all the coefficients from
the fit object and to specify them inside the garchSpec function). So
there is no direct link between the fit method and the simulation one.
Did I miss something ? If not, I clearly prefer the spec >- fit ->
simulation workflow from rugarch.

Question 2 : After fitting the arma-garch models to the various
series, I compute a residulas matrix and I compute the rank of each of
them within the original series. Then I fit a student copula using the
fitCopula function. The next step is to simulate "correlated
trajectories". To do so, I simulate realisations of the student copula
using the rmvdc function, then I directly inject these innovations
inside the garchSim method (this is easy to do by modifying the
function, there is just a parameter to add and a little modification
inside the body to do.)

If I want to do the same job with the rugarch package, how should I
provide to the function the innovations which were previously
generated ?

Thanks for your help,

Ludo



More information about the R-SIG-Finance mailing list