Now I am doing a job of estimating parameters for Archimedean Copulas . I use Matlab as the programming tool to code a MLE function. Here, as a new guy in SAS, I wonder whether there is a existing package of MLE estimation in SAS or I need to code it myself. Thanks

Hi,Have you ever referred to "Fitting copulas to data" , diploma thesis by ROBERTO DE MATTEIS, supervised by Prof. A. McNeil. It can be downloaded. It offers the codes of esitmation and testing using S-plus. Can you use S-plus? It can also be converted to Matlab if you want to. I have done some work on that. You are welcome to discuss the topics with me. Good luck!

S-plus has a commercial module available which contains the MLE estimation procedures for about a dozen different types of copulae. I have not used it myself, I actually built most of the routines from scratch in s-plus, which is not that hard to do. However, this thing could probably save you some serious time. I believe it is called Finmetrics, and contains quite a bit of other functionality in addition to the copula stuff. I don't know about SAS, but I bet they have something comparable.

Wow, people are working so quickly and hard, and they have already worked out the packages for that. I did not find that before and has been working on that for a time. So it is really a bit waste of time. I dont know whether SAS has the corresponding package.

JabairuStork and sunnyseasons, thank you for the information.I have download the paper of ROBERTO. Since I know nothing about S-Plus and the time limit for my project, I will use Matlab to code it. I am estimating Archimedean copulas parameters and it is a one parameter estimation. I plan to extend it to multi-parameter estimation , which is for estimating linear combinations of different copulas. Sunnyseasons, do you have any idea about choosing initial parameters in estimation of copulas?

I have just completed the parameter estimation for the student's t copula with Matlab.I can tell you that the final result is heavily dependent on the quality of initial data. The MLE (or, better, as u might know, CML or IFM) is easy to programme once u know the expression for the log likelihood of the copula density. Doing it for the multivariate case is straightforward. To this end, u can find the expression for the Clayton multivariate copula in Meneguzzo-Vecchiato (2002). An alfa about 0.5 could be a good indicator of your estimation.. for the t copula, the degrees of freedom ranges is about 7-15 DoF..but again don't trust the result immediatly, cause you have to work on the data first and be very careful on missing data and observation frequency..to this end have a look at Mashal-Zeevi paper.Let me know if any problem arises..Now I am comparing CDO prices by using different copulae..Any discussion on it is really welcome..rgds,S.

Does anyone know why the family of Archimedean copulas has the name of "Archimdean"? I did not find a place that says this. Very curious of this....

QuoteOriginally posted by: StefanoneI have just completed the parameter estimation for the student's t copula with Matlab.I can tell you that the final result is heavily dependent on the quality of initial data. The MLE (or, better, as u might know, CML or IFM) is easy to programme once u know the expression for the log likelihood of the copula density. Doing it for the multivariate case is straightforward. To this end, u can find the expression for the Clayton multivariate copula in Meneguzzo-Vecchiato (2002). An alfa about 0.5 could be a good indicator of your estimation.. for the t copula, the degrees of freedom ranges is about 7-15 DoF..but again don't trust the result immediatly, cause you have to work on the data first and be very careful on missing data and observation frequency..to this end have a look at Mashal-Zeevi paper.Let me know if any problem arises..Now I am comparing CDO prices by using different copulae..Any discussion on it is really welcome..rgds,S. My data has negative dependence so using the new software from S-plus (finmetrics) is not working because it does not implement negative dependence yet. I want to fit frank scopula to 2 rv. Can you provide a step by step procedure in Matlab. I have very little time to learn matlab.

Hello,I am a student in Romania at the Academy of Economic Studies Bucharest. I am preparing my gradution paper on the estimation of VaR an EVT, copulas aproach. I have read the paper of ROBERTO DE MATTEIS. Unfortunately I don't have acces to S-plus, however I have MatLab R14. I would like to ask you if it were posible to send me your converted routines from S-plus to MatLab.Thank you very muchQuoteOriginally posted by: sunnyseasonsHi,Have you ever referred to "Fitting copulas to data" , diploma thesis by ROBERTO DE MATTEIS, supervised by Prof. A. McNeil. It can be downloaded. It offers the codes of esitmation and testing using S-plus. Can you use S-plus? It can also be converted to Matlab if you want to. I have done some work on that. You are welcome to discuss the topics with me. Good luck!

Hi Milton,You can get a free student copy of S-plus from S-plus. HTH.CD

I have submited a request and I am waiting for a reply. I couldn't register on the e-academia page but I sent a mail to the contact e-mail. They said they will get back to me in 2 bussines days. I am waiting.Thanks