Message-ID: <1078880.1075863444362.JavaMail.evans@thyme>
Date: Sun, 19 Aug 2001 10:32:58 -0700 (PDT)
From: duffie@smtp1.stanford.edu
To: j.kaminski@enron.com
Subject: Re: The first document for audit
Cc: duffie@smtp1.stanford.edu
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Bcc: duffie@smtp1.stanford.edu
X-From: J D Duffie <duffie@smtp1.stanford.edu>@ENRON <IMCEANOTES-J+20D+20Duffie+20+3Cduffie+40smtp1+2Estanford+2Eedu+3E+40ENRON@ENRON.com>
X-To: Kaminski, Vince J </O=ENRON/OU=NA/CN=RECIPIENTS/CN=VKAMINS>
X-cc: duffie@smtp1.stanford.edu
X-bcc: 
X-Folder: \VKAMINS (Non-Privileged)\Kaminski, Vince J\Inbox
X-Origin: Kaminski-V
X-FileName: VKAMINS (Non-Privileged).pst

Vince:

I read the relibaility-based VAR method document with interest.

Here are some comments:

1. The document falls short of describing how
to use the method in practice, but rather
gives some general motivation. I am confident
that I myself could not develop an algorithm
based on this document, unless all of the
derivative valuation functions are explicit
(and rather nice) functions
with explicit derivatives of the underlying variables, in which case
the VAR problem is not so hard in any case,
as Monte Carlo is very effective in this case.
(My comment here is critical of exposition.
I do not suggest that the proposed method
is weak!)

2. The remark (p. 2) that my paper with Jun Pan
addresses only normal returns is not
correct. The main point of our paper was to allow
for jumps, including fat tails and skewness.
The comment that we rely on delta-gamma
is a good criticism,and the main potential
advantage of the reliability approach,
if it works in practical cases.

3. There is a typo at page 2, at 9 lines from
the bottom, where I presume g(T(u)), not g(T(x))

4. There is a missing `we' on page 5, 9 lines
from the bottom.

5. The remark on CPU time (page 7) does
not mention what machine was used.

6. The example (page 8) does
not really have much non-normality. Skewness=0,
and kurtosis (4) (is this excess kurtosis?)
are not much compared to most practical cases, in which kurtosis
is an extra order of magnitude, with lots
of skewness.


Overall, the paper piqued my curiosity a lot,
and left me wanting more clarity,
and a cookbook.


I hope `events' at Enron have not made things
difficult for you.

I will later bill you for one hour, if that is OK.


Warmest personal regards, Darrell



> X-MimeOLE: Produced By Microsoft Exchange V6.0.4418.65
> content-class: urn:content-classes:message
> MIME-Version: 1.0
> Subject: The first document for audit
> Date: Mon, 16 Jul 2001 12:47:45 -0500
> X-MS-Has-Attach:
> X-MS-TNEF-Correlator:
> Thread-Topic: The first document for audit
> Thread-Index: AcEOH2t+YrCeSfbQQ9uW1MfwNSwcgg==
> From: "Kaminski, Vince J" <Vince.J.Kaminski@enron.com>
> To: <duffie@Stanford.EDU>
> Cc: <vkaminski@aol.com>
> X-OriginalArrivalTime: 16 Jul 2001 17:48:09.0130 (UTC)
FILETIME=[79ADE8A0:01C10E1F]
> X-UIDL: 1c3d3038b819e9d553dd6b2f57734529
>
> Darrell,
>
> I am sending you the first document for review. It's a new algorithm
> invented by Rabi De and Tanya.
>
> The credit paper has been written but it needs more work
> before we can send it to you.
>
> I think you will enjoy  the document I am attaching.
>
>
> Vince
>
>
>  <<Reliabiliy_VAR8.doc>>

_____________________________________________
Darrell Duffie
mail  GSB Stanford CA 94305-5015 USA
phone 650 723 1976
fax   650 725 7979
email duffie@stanford.edu
web   http://www.stanford.edu/~duffie/
_____________________________________________