Re: [Quantlib-dev] processes calibration

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

I think there is now a new class called EquityModelHelper according to Klaus' email. I think it has something to do with calibration but I never tried it.

Yee Man

PS My email exchange with Klaus
-------------------------------

----- original message --------

Subject: Re: GJR-GARCH pricing engine and QuantLib
Sent: Tue, 22 Apr 2008
From: Yee Man Chan<ym...@ya...>

>
> --- Klaus Spanderen <kl...@sp...> wrote:
>
> > Hi
> >
> > I've just looked onto the problem and have seen that
> > the HestonModelHelper
> > will also work with the GJRGARCH engine w/o any
> > change (in fact for the next
> > release I'll rename it into EquityModelHelper). The
> > only small change I made
> > was on the GJRGARCH model and I've applied the
> > Heston Dax calibration to the
> > GJR-GARCH model. Please find the new test case
> > enclosed in the attached files
> >
>
> That's great news. I think I can try it out some time.
>
> > Two question remain
> > 1.You set
> > const Real v0 = omega/(1.0-m1)
> > but I guess v0 is a free parameter of the model,
> > right (no constraint needed
> > during the calibration for v0)?
>
> Yeah this is a free parameter that is to be estimated.
> The Duan et al (2006) paper simply set v0 to the
> long-term vol which is omega/(1-m1)
>
> > 2. Are the parameter constraints
> >        arguments_[0] =
> > ConstantParameter(process->omega(),
> >                                         
> > PositiveConstraint());
> >         arguments_[1] =
> > ConstantParameter(process->alpha(),
> >                                         
> > PositiveConstraint());
> >         arguments_[2] =
> > ConstantParameter(process->beta(),
> >                                         
> > PositiveConstraint());
> >         arguments_[3] =
> > ConstantParameter(process->gamma(),
> >                                         
> > BoundaryConstraint(-1.0, 1.0));
> >         arguments_[4] =
> > ConstantParameter(process->lambda(),
> >                                         
> > BoundaryConstraint(-1.0, 1.0));
> >         arguments_[5] =
> > ConstantParameter(process->v0(),
> >                                         
> > PositiveConstraint());
> > plus the Volatility Constraint correct and all we
> > need?
> >
>
> Oops. I think the constraints are not quite right.
> They should be:
>
> omega - > 0
> alpha, beta - >= 0 and < 1
> beta+gamma >= 0 which implies gamma > -1 and < 1
> v0 - > 0
> lambda should have no constraints but my experience is
> that it seems to be between -0.1 to 0.1
>
> You can refer to the GJR (1993) paper for reference.
>
> Regards,
> Yee Man
>
> > cheers
> >  Klaus
> >
> >
> > On Saturday 19 April 2008 23:06:12 you wrote:
> > > Great! So how do I get started implementing the
> > > calibration helper? I think I need to read some
> > > documentation or paper.
> > >
> > > Yee Man
> > >
> > > --- Klaus Spanderen <kl...@sp...> wrote:
> > > > Hi Yee Man,
> > > >
> > > > thanks it's in. Makes it easier for dummies like
> > me
> > > > to use the pricing engines
> > > > properly.
> > > >
> > > > cheers
> > > >
> > > > On Friday 18 April 2008 03:22:01 you wrote:
> > > > > Hi Klaus
> > > > >
> > > > >    I added a parameter called daysPerYr_ to
> > > >
> > > > GJRGARCH
> > > >
> > > > > process. I think this solves the daily
> > constant
> > > > > problem. I also updated the relevant files and
> > > >
> > > > added
> > > >
> > > > > comments to tell people to set this new
> > variable.
> > > > >
> > > > >    Please take a look and let me know if it is
> > ok.
> > > > >
> > > > > Regards,
> > > > > Yee Man
> > > > >
> > > > > --- Klaus Spanderen <kl...@sp...>
> > wrote:
> > > > > > Hi Yee Man,
> > > > > >
> > > > > > done it's in.
> > >
> > >
> >
> http://quantlib.svn.sourceforge.net/viewvc/quantlib/trunk/QuantLib/
> > >
> > > > > > On Wednesday 16 April 2008 01:54:55 you
> > wrote:
> > > > > > > In the case that you calibrate the model
> > using
> > > >
> > > > the
> > > >
> > > > > > > time series of the underlying, I think you
> > > >
> > > > might
> > > >
> > > > > > want
> > > > > >
> > > > > > > to multiply by 251 (is'nt it 252 as
> > suggested
> > > >
> > > > by
> > > >
> > > > > > the
> > > > > >
> > > > > > > Hull book?)
> > > > > >
> > > > > > Seems that we here in Germany have more
> > holidays
> > > > > > than the rest of the world;-)
> > > > > > Yepp, you are right.
> > > > > >
> > > > > > > instead of 365 as in my code. The problem
> > > > > > > here is that the Duan et al paper I used
> > for
> > > >
> > > > test
> > > >
> > > > > > data
> > > > > >
> > > > > > > assumes 365 trading days. So I suppose
> > maybe
> > > >
> > > > it
> > > >
> > > > > > should
> > > > > >
> > > > > > > be the caller of GJR-GARCH's job to
> > annualize
> > > >
> > > > the
> > > >
> > > > > > > parameters?
> > > > > >
> > > > > > May be we should add some extra comment on
> > this
> > > >
> > > > to
> > > >
> > > > > > the GJRGarchProcess to make
> > > > > > this clear (I as a more "practitioner" would
> > > >
> > > > have
> > > >
> > > > > > expected annualized
> > > > > > parameters, e.g. "my" beta would be 365
> > times
> > > >
> > > > your
> > > >
> > > > > > beta).
> > > > > >
> > > > > > > Do you mean it is very easy to write this
> > > > > > > CalibrationHelper? If so, I think I can do
> > it.
> > > >
> > > > But
> > > >
> > > > > > the
> > > > > >
> > > > > > > problem is I don't know if I can find
> > testing
> > > > > >
> > > > > > data.
> > > > > >
> > > > > >
> > > > > > We need something like the
> > HestonModelHelper.
> > > > > >
> > > > > > Could it be that this model wasn't
> > calibrated to
> > > >
> > > > a
> > > >
> > > > > > real volatility surface at
> > > > > > all in the literatur so far? But the real
> > > >
> > > > problem is
> > > >
> > > > > > the pricing engine and
> > > > > > that you've done already. The rest is not a
> > big
> > > > > > issue. Should we do it
> > > > > > together?
> > > > > >
> > > > > > cheers
> > > > > >  Klaus
> > > > > > --
> > > > > > Klaus Spanderen
> > > > > > Ludwig Erhard Str. 12
> > > > > > 48734 Reken (Germany)
> > > > > > EMail: kl...@NO... (remove
> > NOSPAM
> > > >
> > > > from
> > > >
> > > > > > the address)
> > >
> > >
--- On Thu, 7/10/08, Sylvain Bertrand <syl...@gm...> wrote:

> From: Sylvain Bertrand <syl...@gm...>
> Subject: Re: [Quantlib-dev] processes calibration
> To: ym...@ya...
> Cc: qua...@li...
> Date: Thursday, July 10, 2008, 11:06 AM
> I couldn't find the calibration part in SVN Head's
> "hestonprocess.cpp".
> 
> Is it commited yet or did I just look too quickly at the
> code?
> 
> Regards,
> Sylvain
> 
> 
> On 7/10/08, Yee Man Chan <ym...@ya...> wrote:
> >
> >
> >
> > During my email exchange with Klaus Spanderen, he told
> me he extended the
> > Heston process calibration code to work with other
> processes. He said the
> > new code works with my GJR-GARCH process. I have never
> tested it myself
> > however.
> >
> > Yee Man
> > --- On Thu, 7/10/08, Sylvain Bertrand
> <syl...@gm...>
> > wrote:
> >
> > > From: Sylvain Bertrand
> <syl...@gm...>
> > > Subject: [Quantlib-dev] processes calibration
> > > To: qua...@li...
> > > Date: Thursday, July 10, 2008, 10:26 AM
> > > Hi everyone,
> > >
> > > I was wondering if there was currently a task in
> progress
> > > to implement
> > > calibration for the processes?
> > >
> > > Or is it something that is voluntarily left to
> the quantlib
> > > user?
> > >
> > > Thanks
> > >
> > >
> >
> Sylvain-------------------------------------------------------------------------
> > > Sponsored by: SourceForge.net Community Choice
> Awards: VOTE
> > > NOW!
> > > Studies have shown that voting for your favorite
> open
> > > source project,
> > > along with a healthy diet, reduces your potential
> for
> > > chronic lameness
> > > and boredom. Vote Now at
> > >
> >
> http://www.sourceforge.net/community/cca08_______________________________________________
> > > QuantLib-dev mailing list
> > > Qua...@li...
> > >
> https://lists.sourceforge.net/lists/listinfo/quantlib-dev
> >
> >
> >
> >