Regarding ql/experimental/credit/lossdistribution.hpp

[N_Lassesen]

The function Real probabilityOfNEvents(Size n, const std::vector<Real>& prob) is numerically unstable. This document:
http://www.cs.toronto.edu/pub/reports/na/ma-07-phd.pdf
describes a recursive procedure (formula 2.1) which is numerically stable, and which gives rise to an identical loss distribution. I have been able to replicate the results in Hull and White (2004) using this procedure with an error of only a few basis points.

Rgds
Nicolai Lassesen

[Roland Lichters-2]

Hi Nicolai,

many thanks for the hint. I will have a look at the procedure as soon as I find some spare time. 
Did you also reproduce the CDO examples in Hull and White's paper? Luigi will soon commit the CDO code that uses the loss distribution bucketing algorithm, also to appear in experimental, I assume. We agreed that it will be released latest with version 1.0. The associated test cases compare against the paper's result. I would be very interested in discussing these and any improvements you have in mind.

Best,
Roland

On Mon, Jul 7, 2008 at 9:52 PM, N_Lassesen <[hidden email]> wrote:

The function Real probabilityOfNEvents(Size n, const std::vector<Real>& prob)
is numerically unstable. This document:
http://www.cs.toronto.edu/pub/reports/na/ma-07-phd.pdf
describes a recursive procedure (formula 2.1) which is numerically stable,
and which gives rise to an identical loss distribution. I have been able to
replicate the results in Hull and White (2004) using this procedure with an
error of only a few basis points.

Rgds
Nicolai Lassesen

--
View this message in context: http://www.nabble.com/Regarding-ql-experimental-credit-lossdistribution.hpp-tp18325181p18325181.html
Sent from the quantlib-dev mailing list archive at Nabble.com.


-------------------------------------------------------------------------
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
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev

-------------------------------------------------------------------------
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
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev

[Roland Lichters-2]

Hi Luigi, Nicolai,

I am enclosing a replacement for lossdistribution.xpp in the experimental/credit folder. Finally I had a look at the thesis that Nicolai has pointed to and changed the method probabilityOfNEvents accordingly. I hope I have cited the thesis properly in the comment, please let me know.

Regards,

Roland

---------- Forwarded message ----------
From: Roland Lichters <[hidden email]>
Date: Tue, Jul 8, 2008 at 10:48 PM
Subject: Re: [Quantlib-dev] Re garding ql/experimental/credit/lossdistribution.hpp
To: N_Lassesen <[hidden email]>
Cc: [hidden email]


Hi Nicolai,

many thanks for the hint. I will have a look at the procedure as soon as I find some spare time. 
Did you also reproduce the CDO examples in Hull and White's paper? Luigi will soon commit the CDO code that uses the loss distribution bucketing algorithm, also to appear in experimental, I assume. We agreed that it will be released latest with version 1.0. The associated test cases compare against the paper's result. I would be very interested in discussing these and any improvements you have in mind.

Best,
Roland

On Mon, Jul 7, 2008 at 9:52 PM, N_Lassesen <[hidden email]> wrote:

The function Real probabilityOfNEvents(Size n, const std::vector<Real>& prob)
is numerically unstable. This document:
http://www.cs.toronto.edu/pub/reports/na/ma-07-phd.pdf
describes a recursive procedure (formula 2.1) which is numerically stable,
and which gives rise to an identical loss distribution. I have been able to
replicate the results in Hull and White (2004) using this procedure with an
error of only a few basis points.

Rgds
Nicolai Lassesen

--
View this message in context: http://www.nabble.com/Regarding-ql-experimental-credit-lossdistribution.hpp-tp18325181p18325181.html
Sent from the quantlib-dev mailing list archive at Nabble.com.


-------------------------------------------------------------------------
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
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev

-------------------------------------------------------------------------
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-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev