Christian Fries and I have a new approach to Greeks in the LMM and I'd be interested in comments.You can get the paper from www.markjoshi.com

read the paper. good stuff. I think there are typos in eq (2) (should be t_i not T_i)?A couple of commentsFor digitals/TARNs, have you compared the method to more simple-minded alternatives that banks tend to use like replacing a digital with a collar (and similar methods for TARNs, see eg my paper on TARNs in Wilmott on "sausage" MC, or a similar "stable" method of Pietersz and Regenmortel)Also, specifically for vega, the likelihood ratio between the two schemes (K1 and K0) will involve different vols, right? So, as the number of time steps is increased, the likelihood ratio weights will blow up, is that correct? Basically the two schemes "in the limit" of small time steps cease to be absolutely continuous wrt each other?thanksV

most of the tests we have done are in the paper.re vega, i don't really know why one would want to do many small steps? there may be an issue with one very small step, if you are close to a reset date which we haven't checked. With vega, you have a choice of mean shifting or vol scaling in any case.Although we focus on TARNs in the paper as they are a well-known hard case, it is worth stressing that the methodology is generic, requiring only the choice of a rate to fix at each reset time.

Hi, I am looking for a smart technics of computing greeks for interest rate derivatives such as TARN ( callable) in Libor Market Model with stochastic volatility .First I think that a finite difference technic is not a good approach since I haven't a smooth functions.Second Malliavin calculus cannot handle this type of product in such stochastic model ( I think very dfficult to compute and also numerically).The technic present by Piterbarg in his papers need the simulation of the tangent process.So if I have 20 libors, I need to simulate 20*20 process for the delta, Too time consuming!!! The proxy scheme technic of M Joshi and C Fries seem to be difficult in sthochastic process case since I need to tabulate ( numerically ) density functions.The adjoint method of Glasserman, I dont check it very well but !!!I think Piterbarg , M joshi and others can help me in this situation!!!Thx

Hi, Piterbarg,I'm interested in your article "TARNs: Models, Valuation, Risk Sensitivities". Wonder if you could send me a copy by this e-mail: greenleaves01@gmail.comThan you.

QuoteOriginally posted by: pascal2006The proxy scheme technic of M Joshi and C Fries seem to be difficult in sthochastic process case since I need to tabulate ( numerically ) density functions.Why? The method works for the case of stochastic volatility (I assume you mean stoch vol) in the same way. What do you mean by tabulate? If you write down the Euler scheme of model with a stoch vol then the formula to calculate the transition probability density (and from it the LR) is actually the same as in the non-stoch vol case. Isn't it?

> If you write down the Euler scheme of model with a stoch vol then the formula to calculate the transition probability density (and from> it the LR) is actually the same as in the non-stoch vol case. Isn't it?Absolutely!

QuoteOriginally posted by: pascal2006Hi, The proxy scheme technic of M Joshi and C Fries seem to be difficultThe method of "fries and joshi" is easy. This Joshi--Fries method is news to me.

Hello everybody,What if the initial scheme includes a 'bridging technique'? Is it still possible to adapt easily the proxy-scheme method in this case?More precisely, I am trying to compute the greeks of a cms-spread range accrual in the LMM. Thank you