Re: Question on extrapolation method in term structure classes

Posted by Luigi Ballabio-2 on Apr 13, 2004; 7:20am
URL: http://quantlib.414.s1.nabble.com/Question-on-extrapolation-method-in-term-structure-classes-tp2912p2913.html

Hi Enrico,

On 2004.04.13 13:05, Michelotti Enrico wrote:
> We have a problem in setting the extrapolation method
> in the term structure class (and derived classes) in case
> we use it to calibrate some models.
> It seems to us that it is possible to set this optionality only
> within the inner functions (ex: discountFactor, forward...) and not  
> to an higher level (for example in the constructor).
>
> - Am I right?

Yes, you are.

> - Did you find the same problem?

Yes, in several classes (volatility term structures, interpolations...)  
But I didn't find the time to code a general solution yet.

> - Do you know how to solve this issue?

We should add a couple of enable/disableExtrapolation() methods to  
those classes, setting a corresponding data member. Then, lower-level  
methods would check that data member besides the local argument.
Also, it would probably be more convenient to encapsulate this in a  
basic class and have TermStructure and such inherit from the latter.

> - Do we need to set manually in QuantLib files the extrapolation at
> true as default?

Yes, I'm afraid this is the simplest solution right now.

> - Do you suggest to create another term structure with this
> optionality in the constructor?

No, I'd rather solve this in a more general way. May you file this as a  
feature request on the QuantLib site? (Or if you code it, you can  
submit it as a patch :)

Later,
        Luigi