Re: [Quantlib-dev] Implementing high-low volatility model

[Ferdinando M. Ametrano-3]

Hi Joseph

while I don't have any special contribution on your design I just
wonder why you are considering the effort of adding this framework to
QuantLib. Isn't R a better alternative for this kind of work?

QuantLib focus has always been on derivatives, and while econometric
estimations are often used as input in derivatives models, the two
domains are almost well defined...

ciao -- Nando

On 4/25/06, Joseph Wang <[hidden email]> wrote:

> I'm in the process of trying to implement a class that does a Garman-Klass
> estimation of volatility based on high-low data.
>
> Right now what I have in mind is a class called IntervalQuote(?) that contains
> open, close, high, low information, and this will be used with the TimeSeries
> template to produce
>
> TimeSeries<IntervalQuote>
>
> which will be the input into the calculation classes.  There will be a helper
> class that creates TimeSeries<IntervalQuote> from a vector of dates, open,
> close, high, and low data.
>
> It seems that one should break up the VolatilityModel into two parts.  One
> part is LocalEstimator (?), the second part combinings the daily estimation
> into a time series using constant combining or GARCH, which would be
> subclasses of EstimationCombiner (?)
>
> Since the LocalEstimator will need different types of inputs, there may be a
> need to create a trait.
>
> Thoughts?  I'm especially interested in feedback as to getting the naming
> conventions right.
>
>
>
>
>
> -------------------------------------------------------
> Using Tomcat but need to do more? Need to support web services, security?
> Get stuff done quickly with pre-integrated technology to make your job easier
> Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo
> http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> _______________________________________________
> Quantlib-dev mailing list
> [hidden email]
> https://lists.sourceforge.net/lists/listinfo/quantlib-dev
>

[Joseph Wang]

On Wednesday 26 April 2006 02:47，you wrote：

> while I don't have any special contribution on your design I just
> wonder why you are considering the effort of adding this framework to
> QuantLib. Isn't R a better alternative for this kind of work?

Funny that you should mention this since I am using R for this
work. :-) :-) :-)

I have my R-SWIG interface to QuantLib working well enough so that I can
switch transparently between C++ and R.  For example, it has a type converter
so that an R vector of doubles gets automatically converted to a vector of
doubles which the C++ engine runs on, and then the result gets automatically
converted back to a data.frame which can then be plotted.

Originally, I was planning to do the coding for GARCH and those models in R or
use some package like Rmetrics, but the thing that I found when I started
doing that is that once you have more than about 20 lines of code, you run
into code management and reusability issues.  The good and bad thing with R
is that it is very flexible, by

> QuantLib focus has always been on derivatives, and while econometric
> estimations are often used as input in derivatives models, the two
> domains are almost well defined...
>
> ciao -- Nando
>
> On 4/25/06, Joseph Wang <[hidden email]> wrote:
> > I'm in the process of trying to implement a class that does a
> > Garman-Klass estimation of volatility based on high-low data.
> >
> > Right now what I have in mind is a class called IntervalQuote(?) that
> > contains open, close, high, low information, and this will be used with
> > the TimeSeries template to produce
> >
> > TimeSeries<IntervalQuote>
> >
> > which will be the input into the calculation classes.  There will be a
> > helper class that creates TimeSeries<IntervalQuote> from a vector of
> > dates, open, close, high, and low data.
> >
> > It seems that one should break up the VolatilityModel into two parts.
> > One part is LocalEstimator (?), the second part combinings the daily
> > estimation into a time series using constant combining or GARCH, which
> > would be subclasses of EstimationCombiner (?)
> >
> > Since the LocalEstimator will need different types of inputs, there may
> > be a need to create a trait.
> >
> > Thoughts?  I'm especially interested in feedback as to getting the naming
> > conventions right.
> >
> >
> >
> >
> >
> > -------------------------------------------------------
> > Using Tomcat but need to do more? Need to support web services, security?
> > Get stuff done quickly with pre-integrated technology to make your job
> > easier Download IBM WebSphere Application Server v.1.0.1 based on Apache
> > Geronimo
> > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
> > _______________________________________________
> > Quantlib-dev mailing list
> > [hidden email]
> > https://lists.sourceforge.net/lists/listinfo/quantlib-dev