" /> Coruscation: January 2005 Archives

« December 2004 | Main | February 2005 »

January 31, 2005

Hudson Rendez-vous

From NYC:
?? Dave Rose ??

From Boston:
Exit 5 Rt 89 NW of Concord NH Thursday 3:00pm.

Thursday night:
Thursday evening Rouses Point, NY (Champlain) for the night.

The Anchorage Motor Inn
164 Lake St, Rouses Point, NY 12979
(518) 297-4211

Friday morning:
Highway 15, just north of Montreal

Friday night: Matagami.


Basel default

Probability of Default (PD)
- the probability that a specific customer will default
within the next 12 months.

Loss Given Default (LGD)
- the percentage of each credit facility that will be lost
if the customer defaults.

Exposure at Default (EAD)
- the expected exposure for each credit facility in the
event of a default.

PD, LGD, and EAD are key measures used by Basel II: Peldec.

January 30, 2005

401k planning

401k planning contributions and investing for retirement: now.

Hudson meeting, day 1

Champlain, New York (US/Canada border) to:

Val-d’Or: 365.5 Miles, Estimated 7 hours, 18 minutes
Amos: 407.7 Miles, Estimated 8 hours, 9 minutes
Matagami: 508.7 Miles, Estimated 10 hours,15 minutes

(From MapQuest)

January 29, 2005

How Ratings Agencies Achieve Rating Stability

Surveys on the use of agency credit ratings reveal that some
investors believe that rating agencies are relatively slow in
adjusting their ratings. A well-accepted explanation for this
perception on the timeliness of ratings is the "through-the-cycle"
methodology that agencies use. According to Moody's, through-the-cycle
ratings are stable because they are intended to measure the risk of
default risk over long investment horizons, and because they are
changed only when agencies are confident that observed changes in a
company's risk profile are likely to be permanent. To verify this
explanation, we quantify the impact of the long-term default horizon
and the prudent migration policy on rating stability from the
perspective of an investor - with no desire for rating stability. This
is done by benchmarking agency ratings with a financial ratio-based
(credit scoring) agency-rating prediction model and (credit scoring)
default-prediction models of various time horizons. We also examine
rating migration practices. Final result is a better quantitative
understanding of the through-the-cycle methodology.

By varying the time horizon in the estimation of default-prediction
models, we search for a best match with the agency-rating prediction
model. Consistent with the agencies' stated objectives, we conclude
that agency ratings are focused on the long term. In contrast to
one-year default prediction models, agency ratings place less weight
on short-term indicators of credit quality.

We also demonstrate that the focus of agencies on long investment
horizons explains only part of the relative stability of agency
ratings. The other aspect of through-the-cycle rating methodology -
agency rating-migration policy - is an even more important factor
underlying the stability of agency ratings. We find that rating
migrations are triggered when the difference between the actual agency
rating and the model predicted rating exceeds a certain threshold
level. When rating migrations are triggered, agencies adjust their
ratings only partially, consistent with the known serial dependency of
agency rating migrations.

How Ratings Agencies Achieve Rating Stability

by Edward I. Altman of New York University, and
Herbert A. Rijken of Vrije Universiteit Amsterdam

April 2004.

January 28, 2005

Hudson maps A: overview

STL, MO --------- NYC, NY: 1021 miles 15 hours 42 mins
NYC, NY --- Radisson, Que: 1264 miles 19 hours 26 mins

(NYC, NY -- Bennington, VT: 192 miles 3 hours, 42 mins
Radisson is 650% farther north than Vermont)

Champlain, New York (US/Canada border) to:

Val-d’Or: 365.5 Miles, Estimated 7 hours, 18 minutes
Amos: 407.7 Miles, Estimated 8 hours, 9 minutes
Matagami: 508.7 Miles, Estimated 10 hours,15 minutes

Bear in mind that from Milford NH to Champlain NY is 235 miles,
Estimated 4 hours 21 minutes, while the first day may be more for
others. We will stop along the way and I'd bet no less than a half
hour will be spent at the meeting place in NY/Canada.

The Next Day to Radisson according to AAA:

Matagami: 372 miles, Estimated 8 hours 24 minutes
Amos: 485 miles, Estimated 11 hours 1 minute
Val d'Or: 538 miles, Estimated 12 hours 13 minutes

The ideal is Val d'Or, as we know, but Amos is about as far north
as we should go on Day 1.

2005 Formula One calendar

2005 Formula One calendar.

1. March 6 Australian GP - Albert Park
2. March 20 Malaysian GP - Sepang
3. April 3 Bahrain GP - Bahrain International Circuit
4. April 24 San Marino GP - Imola
5. May 8 Spanish GP - Catalunya
6. May 22 Monaco GP - Monte Carlo
7. May 29 European GP - Nurburgring
8. June 12 Canadian GP - Montreal
9. June 19 United States GP - Indianapolis
10. July 3 French GP - Magny-Cours
11. July 10 British GP - Silverstone
12. July 24 German GP - Hockenheim
13. July 31 Hungarian GP - Hungaroring
14. August 21 Turkish GP - Istanbul Kurtkoy International Circuit
15. September 4 Italian GP - Monza
16. September 11 Belgian GP - Spa-Francorchamps
17. September 25 Brazilian GP - Interlagos
18. October 9 Japanese GP - Suzuka
19. October 16 Chinese GP - Shanghai International Circuit

January 27, 2005

Discount Factor for Estimating Economic LGD

Banks must measure the loss arising from counterparty default in order
to achieve Advanced-IRB compliance under the proposed Basel II minimum
regulatory capital framework. Which discount rate to use on cash
received post-default is a question that is the subject of
considerable disagreement amongst practitioners and banking
supervisors. We review alternative extant proposals and develop a new
method for choosing an appropriate discount rate contingent upon the
risk of the recovery cash flow. An example of how supervisory
determined LGD discount rates could be set is demonstrated.
Empirically, the required rate of return on defaulted corporate bonds
is shown to be similar in magnitude to the yield on BB rated debt. For
defaulted small and medium enterprise (SME) bank loans, the mean
discount rate is found to be similar, on average to the contract rate
pertaining at the time of default.

Choosing the Discount Factor for Estimating Economic LGD
by Ian Maclachlan of Australia and New Zealand Banking Group Ltd.
May 2004

January 26, 2005

Web mathematica takes derivatives.

Web Mathematica takes derivatives.

January 25, 2005


ETFs are shares of a basket of stocks bought and sold as a single
investment. Investment companies create these stocks by buying the
underlying stocks and issuing ETF shares. Unlike mutual funds whose
price is set once per day, ETFs trade on stock exchanges at
constantly changing market prices. This prevents market timers
getting preferential prices like the recent mutual fund scandals.
Very large investors can issue new shares or redeem their shares for
the underlying stocks. This keeps the ETF price close in price to the
underlying shares. ETFs do not trade at sizable discounts or
surpluses to the underlying stocks like closed end mutual funds. If
the ETFs begin to trade with any significant discount or surplus,
large investors will issue new shares or redeem their shares to
eliminate the discount or surplus.


January 24, 2005

Ascential Network

The Ascential Network's industry experience in data integreation
leverages strategic synergies across enterprise platforms.

January 23, 2005

Treeage statistical software for non-statistician

TreeAge offers statistical software for non-statisticians.

Features include sensitivity analysis and distribution graphs.

January 22, 2005

Belief Networks and Decision Networks

Belief networks (also known as Bayesian networks, Bayes networks and
causal probabilistic networks), provide a method to represent
relationships between propositions or variables, even if the
relationships involve uncertainty, unpredictability or imprecision.

They may be learned automatically from data files, created by an
expert, or developed by a combination of the two. They capture
knowledge in a modular form that can be transported from one situation
to another; it is a form people can understand, and which allows a
clear visualization of the relationships involved.

By adding decision variables (things that can be controlled), and
utility variables (things we want to optimize) to the relationships of
a belief network, a decision network (also known as an influence
diagram) is formed. This can be used to find optimal decisions,
control systems, or plans.

Norsys bayesian belief software, based in Vancouver, Canada.

January 21, 2005

Agena Risk bayesian network

Agena Risk bayesian network analysis software and whitepapers.

January 20, 2005

Global Forex Trading

Global Forex Trading offers DealBook® FX 2 'powerful online
currency trading' tool that provides you with instant access
to the forex market.

January 19, 2005


Quinetix offers statistical and optimization consulting.

January 18, 2005


That he and Mr. Begala would be allowed to lob softballs at a man who
may have been a cog in illegal government wrongdoing, on a show produced
by television's self-proclaimed "most trusted" news network, is bad
enough. That almost no one would notice, let alone protest, is a
snapshot of our cultural moment, in which hidden agendas in the
presentation of "news" metastasize daily into a Kafkaesque hall of
mirrors that could drive even the most earnest American into abject
cynicism. But the ugly bigger picture reaches well beyond "Crossfire"
and CNN.


January 17, 2005

Winter driving report

Winter driving report
Location: Sandy Hook Manitoba Ca

Extreme cold temp. report
The over night lows here are hitting -30F and with the windchill it feels like

My MC starts every time,first time,with no block heater.

North American Motoring

Join Date: Mar 2004
Location: Sandy Hook Manitoba Ca
Posts: 3,359
The over night lows here are hitting -30F and with the windchill 
it feels like -45F
But it's a dry cold,humidity is around 40%.
My MC starts every time, first time,with no block heater.I do 
have a trickle charger that plugs into the cig. lighter, keeps the 
battery topped up,and warm.
I make sure I turn off everything at shut down,heater 
fan,radio,and that the e-brake is on,so the DRLs are off. I find it 
cranks alittle slower than normal, but catches everytime. Idle is 
rougher than normal for about 15 seconds. Then it levels out at about 
1200RPM, I hold the clutch in for about 30 seconds or so,the car 
stalled once last winter after I let the clutch out to fast, the 
tranny oil is thick at these temps. Then I slowly let the clutch out 
and keep the idle up till the pedal is fully released, it is a bit 
noisy, but that goes away after things start moving.
Then I turn on the heater fan, seat heat,never use high setting, 
rear defog, and the AC to dry the air to help clear the windows.At 
these temps. I let the car warm up for between 5-10 min.But I have 
driven away once I'm buckled up,and the car just drives away.
Now I do keep revs below 3000RPM until the car is up to normal 
operating temp.
The Pirelli Snowcontrols don't seem to get "square wheel" as bad 
as the all seasons I ran last winter,could be that they run at 44PSI. 
The cockpit is very warm,all the windows are clear,and the little guy 
still goes like snot.
Now there are a few new rattles,but at these temps. I rattle too. 
But with the heater blasting,radio going over that,not to mention 
paying attention to the ice and snow,I don't really hear them.
The MPG tank at these temps. I fill my tank at 1/2 empty 
normally, and I usually go 400KMs by then.At these temps I get between 
300-325KMs at 1/2 a tank.Warm-up,wheel spin,and plowing through snow 
play into this,but the cold temps cause bearings to freeze up,and I'm 
sure that it is harder for the car to run at these temps period.
This car works great at these temps,and I feel safe on my hour 
highway commute in it.

Daou Report

Daou Report ia blog portal with refreshed ledes.
Mostly political; page layout shows left is left and right is right.

January 16, 2005

Texas economic variables

DataBasics walks you through the essentials of Texas economic data.
The articles present numeric operations that economists use to make
data more meaningful. The data definitions are descriptions of
frequently used Texas economic variables that, taken together, help
paint a picture of statewide economic activity.

Texas economic variables.

January 15, 2005

BLAST and SAS: String matching algorithms and their application

Was the author really Shakespeare?- String matching algorithms and
their application
Raymond Wan
Gilead Sciences, Inc.

Sting matching especially approximate (fuzzy) string matching is
important to a lot of different fields in computer science. It is
used in CRM, database cleaning, bibliometrics, and especially
bioinformatics. In fact, a large portion of the supercomputing
resources in the world is now devoted to an algorithm called BLAST
(Basic Local Alignment Search Tool), which is a fuzzy string matching
algorithm. At the heart of all these applications is the need to
measure how different two text strings are to each other. We will look
at two different ways to build a measure and how it can be
implemented in SAS.

June 2004 Bay Area SAS Users Group Meeting

When: 8 June 2004
Time: 1:00-5:00 p.m.
Where: Gilead Sciences, Inc.
320 Lakeside Drive, Foster City, CA 94404

Host: Steve Wong

Glenn Itano (650-522-5671)
Sandy Chang (650-522-5285)

January 14, 2005

Bayesian Methods for Improving Credit Scoring Models

Abstract: We propose a Bayesian methodology that enables banks with
small datasets to improve their default probability estimates by
imposing prior information. As prior information, we use coefficients
from credit scoring models estimated on other datasets. Through
simulations, we explore the default prediction power of three Bayesian
estimators in three different scenarios and find that all three
perform better than standard maximum likelihood estimates. We
therefore recommend that banks consider Bayesian estimation for
internal and regulatory default prediction models.

Keywords: Credit Ratings, Rating Agency, Bayesian Inference, Basel II

JEL Classification: C11, G21, G33

Bayesian Methods for Improving Credit Scoring Models

by Gunter Löffler of the University of Ulm,
Peter N. Posch of the University of Ulm, and
Christiane Schoene of the University of Ulm

Posted 2004 December 16.

January 13, 2005

Receiver Operating Characteristic (ROC)


The ability of a test to discriminate diseased cases from normal cases
is evaluated using Receiver Operating Characteristic (ROC) curve
analysis (Metz, 1978; Zweig & Campbell, 1993). ROC curves can also be
used to compare the diagnostic performance of two or more laboratory or
diagnostic tests (Griner et al., 1981).

January 12, 2005

Lindeberg's central limit theorem

Lindeberg's Central Limit Theorem at Planetmath.

January 11, 2005

White Collar Crime

Law professors' white collar crime covers the legal issues in business crime.

White Collar Crime also covers the public out of courtroom contests.

Put the search words "Enron scandal" or "Ken Lay," or even this Enron
reporter's name, "Mary Flood," into any of the above search engines and
one of the first things you will see is www.kenlayinfo.com. If you hit
on Lay's Web site from there, then Lay pays between roughly 5 cents and
12 cents.

January 10, 2005

In the Agora

In the Agora is a nice name for an Indianapolis-centric current
events blog by Zach Wendling and Paul Musgrave.

Recent controversies: Daylight Savings time.

January 9, 2005

Mortgage Blogs

The Mortgages Blog (by weblogsinc) is frequently updated with
mortgage industry news mixes some news items and more outside
links with descriptions.

Bankrate is potpourri of mortgage and consumer finance content,
more written for consumers than lenders, and syndicated onto
many consumer sites.

January 8, 2005

Clusty headlines

Clusty headlines are groupable into clusters
by reader-specified criterion.

Clusty shows Stylized Facts as in these clusters:

Market (30)
⇨Growth (17)
⇨Statistical, Empirical (10)
⇨Interest Rates (9)
⇨Bank, Research (7)
⇨Volatility, Modeling (7)
⇨Behavior, Generate (6)
⇨Generate The Stylized Facts (5)
⇨Economic Blog (3)

Building a Smarter Search Engine"
Business Week (01/04/05); Green, Heather

Search engine startup Vivisimo uses artificial intelligence technology
developed at Carnegie Mellon University to organize search results
more neatly. The company launched a consumer metasearch service called
Clusty three months ago that pulls results from other search engines,
then uses its own technology to cluster those results according to
major themes. A search for "seal" will produce the normal mix of
references in the main results--including pages referring to the sea
mammal, Seal the musician, Wet Seal-brand clothing, and Navy SEALS--as
well as folders on each of those topics on the left-hand side. By
opening the appropriate folder, users will be able to drill down to
find more relevant results. Clusty is a good tool for managing a
growing amount of online information and makes search technology
easier to use, says Search Engine Showdown operator Greg Notess.
Analysts say Vivisimo faces an enormous challenge in the general
search space as much larger companies such as Google, Microsoft, and
Yahoo roll out more features, such as desktop and multimedia search.

Although Vivisimo has been profitable for the past two years, it will
need to focus on niche search markets in order to grow in the consumer
search space. Vivisimo's unique search technology could be used to
produce superior results for health care, travel, or other
industry-specific searches. Founder Raul Valdes-Perez, who started
Vivisimo with two other CMU researchers and a $1 million grant from
the National Science Foundation, says Clusty will gain market share
because it makes Web searching easier. Vivisimo also provides
enterprise search tools and licenses its technology to popular
consumer Web sites.

January 7, 2005

TreeBoost - Stochastic Gradient Boosting

TreeBoost - Stochastic Gradient Boosting.

"Boosting" is a technique for improving the accuracy of a predictive
function by applying the function repeatedly in a series and combining
the output of each function with weighting so that the total error of
the prediction is minimized. In many cases, the predictive accuracy of
such a series greatly exceeds the accuracy of the base function used

Hudson Trip info 2: clothes

See also MINI2 thread, New England MINI thread and Ducttape Dave Rose's master site.

*** What to bring ***

1. Clothing: [US Elite gear]

1.1 N3B 'snorkel' Parka

1.2 gloves

1.3 socks

1.4 pants:
F-1B Extreme Cold Weather Trousers
F1B Flight Pants from Army Navy Sales.
or Military ECWCS Hyvat Trousers at Army navy Deals
or fleece bear pants.

January 6, 2005

Info 1: The MINI

See also MINI2 thread, 1, 2; New England MINI thread and Ducttape Dave Rose's master site.

*** What to bring ***

1. Snow Tires

2. tow eye

3. tow chain

4. batteries

5. Fuses

6. Maps:
North America map
western Quebec map

7. FRS radio

8. Battery chargers

9. Maps

10. GPS receiver

11. Ice window scraper

11. paper towels

12. Window cleaning de-icing juice

13. CDs to share

14. Traction sand

15. Flare/light

16. Snowshoes.

17. Tarp

PRMIA's risk news

PRMIA's risk news.

Week In Risk is published every Friday by Risk Communications.
It is (was ?) distributed by PRMIA, the Professional Risk Managers
International Association.

January 5, 2005

Market-generated forecasts are typically accurate

Prediction Markets

We analyze the extent to which simple markets can be used to aggregate
disperse information into efficient forecasts of uncertain future
events. Drawing together data from a range of prediction contexts, we
show that market-generated forecasts are typically fairly accurate,
and that they outperform most moderately sophisticated benchmarks.

Carefully designed contracts can yield insight into the market's
expectations about probabilities, means and medians, and also
uncertainty about these parameters. Moreover, conditional markets can
effectively reveal the market's beliefs about regression coefficients,
although we still have the usual problem of disentangling correlation
from causation. We discuss a number of market design issues and
highlight domains in which prediction markets are most likely to be

Justin Wolfers (Stanford University)


January 4, 2005

Credit Spread determined by a single common variable

Understanding Credit Spread Markets

I describe results of empirical studies of historical spreads in the
corporate bond market and show how spread changes, regardless of
credit rating, are largely determined by a single common variable. In
addition, I calculate the component of corporate credit spreads due to
default probability and report analyses of the residual spreads as
functions of credit quality and duration.

Analysis of spread changes for other spread markets (e.g.
asset-backeds, emerging markets, etc.) reveal a large, but smaller,
degree of spread co-movement across sectors and indicate that sector
spreads trend and mean revert on roughly similar time scales. That is,
spread changes trend in the short term (under two years) and
mean-revert over longer periods.

That information, along with the CAPM and our strategists' monthly
outlooks, was used to construct a cross-sector asset allocation model
that consistently outperformed a benchmark portfolio in ten years of
out-of-sample testing.

Terry Benzschawel (Citigroup Global Markets)


January 3, 2005

FIASI: Fixed Income Analysts Society

Fixed Income Analysts Society (FIASI) director bios.

January 2, 2005

Cold Mini


MINI tests cars in Finland.

[Via Motoring File]

Correlation Monger

Correlation monger provides pair-wise correlation of
demographic variables across 50 US states. For example,
Canadians increase property values.

January 1, 2005

Credit scoring overdose

As the director of risk analysis for the Office of the Comptroller of
the Currency, Jeffrey Brown spends most of his time reviewing how
banks use credit scoring and other risk-management tools. [1]

Credit scoring has revolutionized lending, but critics worry
institutions are pushing this technology beyond its limits.

Hudson Index

Hudson index