now Copula is a standard method for modeling CDS/CDO. what are the main problems of the copula? Any better model?Thanks

the most-well known problem about copula is that it is for random variables but not process. i vaguely remembered someone from Europe proved theoretically there always exists a process corresponding to each copula, but he stopped at the existence theorem and did not give an algorithm for how to find hte process (maybe he couldnt or it is simply infeasible). i do not know any better model in this regard, but i am all ears.

The problem is determining which coupula to use. There was a paper that showed the capital requirements for a loan portfolio could increase by as much as 10X depending on the coupula used.

hi, thanks. Do you know the name of guy and whether he publish his work?

Hello. There are several problems with the use of copulas. The following are the remarks made by the academic Thomas Mikosch (one of the world leading figures in stochastic modelling) in his paper Copulas: Tales and Facts. I recomend you to read that paper.-There is no particular advantage of using copulas when dealing with multivariate distributions.Instead one can and should use any multivariate distribution which is suited to theproblem at hand and which can be treated by statistical techniques.- The marginal distributions and the copula of a multivariate distribution are inextricablylinked. The main selling point of the copula technology | separation of the copula (dependencefunction) from the marginal distributions | leads to a biased view of stochasticdependence, in particular when one ts a model to the data.-Various copula models (Archimedean, t-, Gaussian, elliptical, extreme value) are mostlychosen because they are mathematically convenient; the rationale for their applications ismurky.-Copulas are considered as an alternative to Gaussian models in a non-Gaussian world. Sincecopulas generate any distribution the class is too big to be understood and to be useful.-There is little statistical theory for copulas. Sensitivity studies of estimation proceduresand goodness-of-t tests for copulas are unknown. It is unclear whether a good t of thecopula of the data yields a good t to the distribution of the data.-Copulas do not contribute to a better understanding of multivariate extremes.-Copulas do not t into the existing framework of stochastic processes and time series analysis;they are essentially static models and are not useful for modeling dependence throughtime.

hi wallclimberThanks for your reply, appreciate.

Hi everyone,The pamplet of T. Mikosch is subject to caution. Here is a link to a comment on Mikosch's paper that will appear soon in Extremes.It also contains references to research articles on copulas.http://neumann.hec.ca/pages/bruno.remil ... e.pdfBruno

Peter Tankov http://www.math.jussieu.fr/~tankov/ has done work on Levy Copulas. As far as I understand this should help with one of the big problems of using copulas for basket option pricing - how to relate the copula of returns over one time period to that of returns over a longer or shorter period.

Could anyone give me a link to a paper introducing using copulas in finance. Thanks

for beginning, you have the thesis of Andrew Patton (see "research") : http://fmg.lse.ac.uk/~patton/for goodness-if-fit, this little survey is useful : http://www.danielberg.no/publications/CopulaGOF.pdfand the website of Paul Embrechts gives some exemples about using copulas for finance : http://www.math.ethz.ch/~baltes/ftp/papers.html

CactusMan, you'll find some excellent papers athttp://gro.creditlyonnais.fr/content/rd/home_copulas.htm

Thank you!

You will find an excellent paper on Dynamic Copula Processes:A new way of modelling CDO trancheshttp://www.cerna.ensmp.fr/Documents/DTT-MA-Dyn ... .pdfcheers

Thanks.