hi everybody,this might be trivial to many of you..I'm implementing pricing basket options using copulas(recent paper by Dong Qu published in wilmott)..i'm testing the simplest situation..a black scholes world(where cdf's can be calculated explicitly), but the basket prices results are quite far from results calculated by monte carlo simulation...I'm suspecting that i'm doing something wrong in the step where i'm transforming the generated gaussian numbers(after imposing the correlation) to uniform variables...I'm doing this by taking the normal distribution of the generated gaussian numbers(it sounds wrong for me,but i don't have an alternative solution)..any help?

sounds not so bad for me. what are you modelling precisely. the time of default? after having the [0,1] uniform distributed rv with the desired dependencestructure you have to apply the inverse of the marginal cdf to each component of your random vector...

QuoteOriginally posted by: genkideskasounds not so bad for me. what are you modelling precisely. the time of default? after having the [0,1] uniform distributed rv with the desired dependencestructure you have to apply the inverse of the marginal cdf to each component of your random vector...i am planning to use it for equity basket option pricing(i'm not sure if the method is as expensive as normal monte carlo simulation,having to get cdf's for all components, as well as simulating variables)..i am doing what you said concerning taking the inverse of the marginal cdf for each component..but still..

zaher do you have a link to the paper you are talking about as I would be interested to read it. Thanks

meteor,give me ur e-mail and i will send it to you..i'm not sure if by putting it on the forum i will be violating the copyright rules..especially that it's an article from wilmott

I would also be very keen to see this if someone could point me toward a link...Cheers

amazing!! after i learned about copulas maybe a year ago, it struck me that people pricing multifactor equity derivatives had all along been using the gaussian copula without realising it.i posted something on this forum back then, saying that even top quants in equity derivatives groups in major investment banks didn't know much at all about copulas.someone else said that was ridiculous, surely IB quants knew all about copulas and were purposely using the gaussian copula.i insisted that equity derivs people had no idea, and should consider using alternative copulas. the thread petered out.has Dong Qu written an article about this? or am i misunderstanding?