Hull-White calibration problem

[Guowen Han]

Hi, group;

I got some strange issue with the Hull-White analytic formulae calibration with simplex method on swaption volatility.

I finished one year calibration except on the day "2005-03-01", and here is error message:
In function `QuantLib::Real QuantLib::Solver1D<Impl>::solve(const F&, QuantLib::Real, QuantLib::Real, QuantLib::Real, QuantLib::Real) const [with F = QuantLib::JamshidianSwaptionEngine::rStarFinder, Impl = QuantLib::Brent]':
root not bracketed: f[-10,10] -> [8.054229e-05,8.054229e-05]

Could someone tell what is the problem?

Thanks.

Guowen

[Luigi Ballabio]

On 11/22/2005 07:41:23 PM, Guowen Han wrote:
> I got some strange issue with the Hull-White analytic formulae
> calibration with simplex method on swaption volatility.
>
> I finished one year calibration except on the day "2005-03-01", and
> here is error message:
> ...
> root not bracketed: f[-10,10] -> [8.054229e-05,8.054229e-05]
>
> Could someone tell what is the problem?

Hard to say without seeing the data. However, the error above means  
that the solver cannot find a value corresponding to your data.  
Shortly, it is trying to find a value for which its results are  
negative (i.e., below your input datum) and one for which is positive  
(i.e., above the input) so that it can scan the range between the two  
values and find the zero (i.e., the value corresponding to the input.)  
But it cannot find such a bracket, since all it finds are positive  
results.

Later,
        Luigi


----------------------------------------

Within C++, there is a much smaller and cleaner language struggling to  
get out.
-- Bjarne Stroustrup

[Guowen Han]

Hi, Luigi;

I guess the step (lambda) for Simplex method  was too small. After I increase it, it work fine.

//  Simplex simplex(0.05, 1e-10);
Simplex simplex(0.1, 1e-10);

One other thing, is there any big picture of the code design?
I have some basic idea for design patterns, but it is still very hard for me to
understand the code.


Thanks.
Guowen

Luigi Ballabio <[hidden email]>
Sent by: [hidden email]

11/24/2005 03:16 AM

To	Guowen Han <[hidden email]>
cc	[hidden email]
Subject	Re: [Quantlib-users] Hull-White calibration problem

On 11/22/2005 07:41:23 PM, Guowen Han wrote:
> I got some strange issue with the Hull-White analytic formulae
> calibration with simplex method on swaption volatility.
>
> I finished one year calibration except on the day "2005-03-01", and
> here is error message:
> ...
> root not bracketed: f[-10,10] -> [8.054229e-05,8.054229e-05]
>
> Could someone tell what is the problem?

Hard to say without seeing the data. However, the error above means
that the solver cannot find a value corresponding to your data.
Shortly, it is trying to find a value for which its results are
negative (i.e., below your input datum) and one for which is positive
(i.e., above the input) so that it can scan the range between the two
values and find the zero (i.e., the value corresponding to the input.)
But it cannot find such a bracket, since all it finds are positive
results.

Later,
Luigi


----------------------------------------

Within C++, there is a much smaller and cleaner language struggling to
get out.
-- Bjarne Stroustrup



-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems?  Stop!  Download the new AJAX search engine that makes
searching your log files as easy as surfing the  web.  DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_idv37&alloc_id865&op=click
_______________________________________________
Quantlib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users

[Luigi Ballabio]

On 11/29/2005 05:53:20 PM, Guowen Han wrote:
> I guess the step (lambda) for Simplex method  was too small. After I
> increase it, it work fine.
>
> //  Simplex simplex(0.05, 1e-10);
> Simplex simplex(0.1, 1e-10);

Glad to hear it.

> One other thing, is there any big picture of the code design?
> I have some basic idea for design patterns, but it is still very hard
> for me to understand the code.

Yes, I'm aware of the problem. I'm slowly trying to write something.

Later,
        Luigi


----------------------------------------

Brady's First Law of Problem Solving:
        When confronted by a difficult problem, you can solve it more
        easily by reducing it to the question, "How would the Lone
        Ranger have handled this?"