Hi Nonius, I am currently working on a model that is very similar to the model you were talking about. At least ZCs look like exp(<A,X> + B). where X is an affine process (unfortunately CIR and OU components a mixed...)Before coming to implementation I feel a bit afraid of pricing long swaptions. 10 ZC terms power 7 give too many monomials. Have you overcome this problem? Do it really work as fast as it is claimed?BTW, did you try to price somthing like (a + b * SwapRate1) 1_{c + d * SwaptRate2}?Thank you.

QuoteOriginally posted by: tourkineHi Nonius, I am currently working on a model that is very similar to the model you were talking about. At least ZCs look like exp(<A,X> + B). where X is an affine process (unfortunately CIR and OU components a mixed...)Before coming to implementation I feel a bit afraid of pricing long swaptions. 10 ZC terms power 7 give too many monomials. Have you overcome this problem? Do it really work as fast as it is claimed?BTW, did you try to price somthing like (a + b * SwapRate1) 1_{c + d * SwaptRate2}?Thank you.Sorr, this is way belated, and I don't know if you even log on anymore, but....yes, if one is not careful, it will be way too slow. actually, I am pretty sure that Goldstein et al simply did some sort of extrapolation to get the higher moments. actually, I think they set the 6th and 7th cumulant to ZERO in their expansion out to 7 moments. That is to say, and they have a footnote concerning this...they approximate out to seven moments by using an approximation of the exp function out to seven terms, then, they get a solution, then, they set the sixth and seventh moment to zero.I am not clear on the most combinatorially efficient way to build those moments, other than to do it recursively.