Problem: Calibration of the Duffie Pan and Singleton Model (2000)

Questions
, , ,
1

Hello community,

at the moment I try to implement the calibration scheme described by (Mikhailov, Nögel).
I’m not sure whether to use NMinimize, ArgMin, FindFit, LeastSquares etc. to solve my calibration problem.
So far I tried NMinimize but the only results I get are error messages (as shown in the image attached).
My question is, how do I need to build my function that NMinimize solves for the five parameters of the model if that is even possible?

Thank you in advance!

Calib.JPG
Calib.JPG1004×564 72.7 KB

2

That’s interessant that you post it, I have exactly the same problem!

  1. How to use exactly the same Squared Error loss function (Link:Estimation process/Calibration ) to achieve the goal of minimizing Modellprices and Marketprice difference and to achieve an approximation to the marketprices curve.
    “NMinimize” only gives me my initial parametervalues back, but not optimizing them as I would like to see, as the theory is explaining…

“FindFit” sounds like a good solution:

FindFit.jpg
FindFit.jpg856×156 44.5 KB

  1. How to loop the HestonCallMod a certain amount of times to achieve this approximation and to get results for our five dimensional parameterset (?(vol of variance);?(mean of long run variance; ?(mean reversion speed);V(0)(initial variance;?(correlation))

“Do” could workout somehow… but how^^ I am trying days now…

Philipp, thanks for starting this discussion round. Calibration and optimization is kind of tricky!

3

philipp, The NMinimize:ivar warning appears to say that k (kappa) has a value and should not be used as an independent variable in NMinimize.

Both: attaching a notebook file with your code would make it easier for others to try to help.

4

Hey Bruce,

thanks for looking it up. I just uploaded the nb.

Best regards,
Pantelis

5

Hey Bruce,

thanks for your fast answer!
I just uploaded my complete notebook file.

Best regards, Philipp