You simulate the short rate under your HW model; then the spot CMS rate is just a function of that short rate (that you can express in terms of zero-coupon bonds, which are analytic under HW given the short rate).
Thanks, mghiggins! What you said is applicable to each path in MC simulation. How about lattice? Starting from H-W, you can build a binary tree of short rates, how do I get CMS 10Y on each node?
You have analytic formula for zero-coupon bond as function of the short-rate : P(t,T) = A(t,T) exp(-B(t,T)*r(t) )and CMS 10y (t) = [ P(t,t_0) - P(t, t_0+10Y) ] / A(t)with A(t) be the annuity
Well, you could try a 2-factor Hull-White model (G2++).Note that there are very good (and easy to implement) approximations for options on CMS-rates given underlying n-factor Gaussian models, see http://ssrn.com/abstract=1093710