Serving the Quantitative Finance Community
Black Karasinski
In the Fincad implementation of the Black Karasinski model d ln(r(t)) = [θ(t) − a(t) ln(r(t))] dt + σ(t)dW(t)which of the parameters θ(t),a(t) and σ(t) are time dependent and which are time independent please?
Black Karasinski
Heavy night's drinking yesterday?The (t) indicates they are time dependent
Black Karasinski
Sorry. Formula is a general one for BK not from Fincad documentation. Just looking to know which of the parameters are actually dependent on t in their implementation. BTW just the usual night's drinking last night!!
Black Karasinski
FINCAD documentation is pretty good, give it a try. It's hard to know what their software is saying so posters here may not know how to offer guidance. I would say that is the canned documentation isn't helpful then call them they are very responsive.In general, time dependent parameters like θ(t) will mean a curve of "instantaneous forwards." So that one really should be time dep.
Black Karasinski
Am working slowly through this. Am now at the stage where I am working with this specification ln(r(t)) = [θ(t) − a ln(r(t))] dt + σdW(t)I have built a trinomial tree and am now trying to calibrate to the swaption volatility surface. I am using the Levenberg Marquardt for this purpose. Does anyone have suggestions regarding the initial values of a and σ (relative to vol of swaptions?) in order for this algorithm to have a fighting chance of working? Thanks.
Black Karasinski
Has any one who has implemented this for Bermudans seen that convergence speed and accuracy is significantly dependent on the underlying shape of the standard swaption term structure?