Re: [Quantlib-users] VAR Methodology

[Debashis D. <dut...@gm...>]

Hi Javit  & All Project Memers,

I have hot received your mail about the riskstats.cpp file you talked about.
Please resend the same to me and other project members who have not received
the same.

While in  VaR, there are  basically two approaches, parametric and
non-parametric.Parametric VaR like VCV VaR requires variance covariance
matrix, where codes can pay a lot making validation rule for making it
positve semi definitive etc. Here many things we can play in code and add
value to real practitioners as well as researchers.

While non-parametric VaR like Historical Simulation and Monte Carlo VaR does
not requires parameters, HS requires data. While Basel II stipulates atleast
1 year data, the ideal is 3 to 5 years data for Historical Simulation as it
seen empirically. But here code can not play much unless we use bootstrapped
Historical Simulation VaR to handle data scarcity.

While I agree with Javit, Monte Carlo a little straight forward to implement
and a good candidate as a VaR  as it tackles path dependency, fat tails,
non-linearity and optionality. But  we need to fix the methodological issues
at the beginning. We must have clarity regarding

1. Whether we shall choose intitally Monte Carl with Single risk factor or
multiple risk factors
2, Which Variance Reduction Techniques

After we freeze these issues , we could move forward.

I think it is a good idea of Jevit  we should share the instruments and then
implement our solutions piece by piece.But I believe before actually kick
start the project, a quick step by step approach regarding existing
development life cycle defining  various pricing engines, processes and the
instrument classes to all project members could be a good beginning.

So I propose initial briefing to all project members about existing
development process with example codes and other briefing regarding
engines.The we fix methodologies and thereafter devolve a distributed
process for developement.

Charles is also back on the project and  he is also a good resource at it as
he  contributed to South Korean Calender  in Quantlib version 0.9.6.

Wish all Project Members the best,

Kind Regards,
Debashis



On 31/08/2008, Cavit Hafizoglu <ch...@vi...> wrote:
>
>  Dear project members,
>
> I'm not sure if you recevied my e-mail about the riskstats.cpp file. In
> case, that is the file where some statistical functions are implemented. It
> would be helpful to read the source code for a beginning.
>
> The next thing is to decide what to implement as a VaR methodology. I am in
> favor of Monte Carlo VaR which I think would be easier to implement, but
> I think we should discuss this further.
>
> Quantlib has a very beatifully structured design. We have to get familiar
> with various pricing engines, processes and the instrument classes. Then,
> the next step is to discuss how to implement VaR. I would recommend reading
> source codes in the test-suite. Maybe, we should share the instruments and
> then implement our solutions piece by piece.
>
> Let me know what you think.
>
> Thank you,
> Javit
>
>
>  On Thu, Aug 28, 2008 at 4:15 AM, Alok Jain <aja...@ya...> wrote:
>
>>
>> That would be a great project. Include me if a project plan is in play. I
>> can throw in my 2 cents.
>>
>> --- On *Wed, 8/27/08, Debashis Dutta <dut...@gm...>* wrote:
>>
>>
>> From: Debashis Dutta <dut...@gm...>
>> Subject: Re: [Quantlib-users] VAR Methodology
>> To: "javit" <ca...@vi...>
>> Cc: qua...@li...
>> Date: Wednesday, August 27, 2008, 12:44 PM
>>
>>
>>  Hi Cavit,
>>
>> Please include me in the project for developing Quantlib in VaR
>> calculation
>>
>> I am very interested in it as I am a little attached to development of VaR
>> Models , as  I   developed and empirically tested different VaR Models
>> including POT model and back testing in Matlab as part of my doctoral
>> dissertation  as well as I am presently involved in developing and
>> implementing Matlab-based VaR models in some banks in GCC region. So it
>> would be an immense learning experience for me.But I need guidance for
>> Quantlib development as I have never developed  anything in Quantlib.
>>
>> I suggest let us first make an implementation plan step wise for
>> developing QuantLib in VaR calculation and then let us  work together. We
>> can also post our progress/queries time to time regarding developent for
>> having advice/suggestions from our  friends  in Quantlib Fraternity.
>>
>> So Let us start and Cheers!
>>
>> Kind Regards,
>> Debashis
>>
>>
>> On 26/08/2008, javit <ca...@vi...> wrote:
>>>
>>>
>>> I'm very interested in developing QuantLib in VaR calculations. Please
>>> feel
>>> free to contact me if you plan on any projects.
>>>
>>> Thank you,
>>> Cavit (Javit) Hafizoglu
>>>
>>>
>>> Ferdinando Ametrano wrote:
>>> >
>>> > On Thu, Aug 14, 2008 at 3:11 PM, ssykowski <sc...@at...>
>>> wrote:
>>> >> We are evaluating different risk libraries for VaR. Can anyone point
>>> me
>>> >> to
>>> >> the details of the approach used by QuantLib.
>>> >
>>> > there is not a proper VAR model in QuantLib, only the ability for a
>>> > statistic accumulator to calculate VAR and assorted tail risk measures
>>> > for a given observed distribution, with and without Gaussian
>>> > assumption
>>> >
>>> > ciao -- Nando
>>> >
>>> >
>>> -------------------------------------------------------------------------
>>> > This SF.Net email is sponsored by the Moblin Your Move Developer's
>>> > challenge
>>> > Build the coolest Linux based applications with Moblin SDK & win great
>>> > prizes
>>> > Grand prize is a trip for two to an Open Source event anywhere in the
>>> > world
>>> > http://moblin-contest.org/redirect.php?banner_id=100&url=/
>>> > _______________________________________________
>>> > QuantLib-users mailing list
>>> > Qua...@li...
>>> > https://lists.sourceforge.net/lists/listinfo/quantlib-users
>>> >
>>> >
>>>
>>> --
>>> View this message in context:
>>> http://www.nabble.com/VAR-Methodology-tp18981245p19165667.html
>>> Sent from the quantlib-users mailing list archive at Nabble.com.
>>>
>>>
>>> -------------------------------------------------------------------------
>>> This SF.Net email is sponsored by the Moblin Your Move Developer's
>>> challenge
>>> Build the coolest Linux based applications with Moblin SDK & win great
>>> prizes
>>> Grand prize is a trip for two to an Open Source event anywhere in the
>>> world
>>> http://moblin-contest.org/redirect.php?banner_id=100&url=/
>>> _______________________________________________
>>> QuantLib-users mailing list
>>> Qua...@li...
>>> https://lists.sourceforge.net/lists/listinfo/quantlib-users
>>>
>>
>>
>>
>> -------------------------------------------------------------------------
>> This SF.Net email is sponsored by the Moblin Your Move Developer's
>> challenge
>> Build the coolest Linux based applications with Moblin SDK & win great
>> prizes
>> Grand prize is a trip for two to an Open Source event anywhere in the worldhttp://moblin-contest.org/redirect.php?banner_id=100&url=/
>>
>> _______________________________________________
>> QuantLib-users mailing lis...@li...https://lists.sourceforge.net/lists/listinfo/quantlib-users
>>
>>
>>
>