Who's sitting on local juries in Louisville, Kentucky?
Nov 10th, 2005 by JTJ

A nice bit of AJ done by the folks at the Louisville
[Kentucky] Courier-Journal
, who analyzed the jury pool and composition
in the C-J's home county.  Some good thinking and moderate
statistical-lifting drives the series.


“Jury not of their peers
In Jefferson County”

People who live in mainly African-American areas
are less likely to serve than those from mostly white areas, a
Courier-Journal analysis found.

Yes, Virginia, methodology DOES matter
Nov 10th, 2005 by JTJ

A piece on calling the elections in Detroit:

MAKING A FORECAST: A secret formula helps producer call the election right



November 10, 2005

What was a viewer to believe?

As polls closed Tuesday, WDIV-TV (Channel 4) declared Freman Hendrix winner of Detroit's mayoral race by 10 percentage points.

WXYZ-TV (Channel 7) showed Hendrix ahead by 4 percentage points, statistically too close to call.

But WJBK-TV (Channel 2) got it right, declaring just after 9 p.m. that
Mayor Kwame Kilpatrick was ahead, 52% to 48%, which turned out to be
almost exactly the final 53%-47% outcome declared many hours later.

And it was vote analyst Tim Kiska who nailed it for WJBK, and for WWJ-AM radio, using counts from 28 of 620 Detroit precincts.

Kiska did it with help from Detroit City Clerk Jackie Currie. She
allowed a crew that Kiska assembled to collect the precinct tallies
shortly after the polls closed at 8 p.m.

Using what he calls a secret formula, Kiska calculated how those 28 precincts would predict the result citywide.

His formula also assumed that absentee voters chose Hendrix over Kilpatrick by a 2-1 ratio.

That's different from the methods of pollsters who got it wrong
Tuesday, Steve Mitchell for WDIV and EPIC/MRA's Ed Sarpolus for WXYZ
and the Free Press. Both men used telephone polls, calling people at
home during the day and evening and asking how they voted.

It's a more standard method of election-day polling, but Tuesday proved treacherous.

Kiska, a former reporter for the Free Press and Detroit News, has done
such election-day predictions since 1974, but said he was nervous

“Every time I go into one of these, my nightmare is I might get it
wrong,” said Kiska, a WWJ producer. “I had a bad feeling about this
going in. I thought there was going to be a Titanic hitting an iceberg
and hoping it wouldn't be me.”

Kiska said he especially felt sorry for his friend Mitchell.

Mitchell said he's been one of the state's most accurate political
pollsters over 20 years, but said his Tuesday survey of 800 voters
turned out to be a bad sample.

He said polling is inherently risky, and that even well-conducted polls
can be wrong one out of 20 times. “I hit number 20 this time.”

For Sarpolus, it's the second Detroit mayoral race that confounded his
polls. He was the only major pollster in 2001 who indicated Gil Hill
would defeat Kilpatrick.

Sarpolus said the pressure to get poll results on the air quickly made
it impossible to adjust his results as real vote totals were made
public during the late evening.

Of Kiska, Sarpolus said: “You have to give him credit. … But you have to assume all city clerks are willing to cooperate.”

Contact CHRIS CHRISTOFF at 517-372-8660 or

Grumbling (again) about only getting half the story
Oct 9th, 2005 by Tom Johnson

long appreciated Ford Fessenden's forceful analytic journalism at the
NYTimes, but a piece he has in today's Week in Review section leaves us
yearning for more. 

In “Where Home Prices Rise Steeply, Bankruptcies Fall,”
Ford raises some interesting — and appropriately inconclusive
questions — about the relationship between real estate prices and the
number of bankruptcies.  And we're given a nicely colored map of
U.S. counties and their changes in bankruptcy rates, 2000 to
2005.  The quartile scale is huge: zero to 35 percent and greater
than 35 percent, both up and down.  The problem is there are no
hard numbers to put the bankruptcies in context related to county
population.  And one or two counties down in southeastern Arizona
have a greater than 35 percent decline in bankruptcies, but we know
they have very sparce populations. 

you might say, “there's simply no room to put all those numbers in the

Right, but they surely could be put online in a
variety of ways.  If there were three bankruptcies in 2005 and two
in 2005, that's pretty close to a 35 percent decline, but hardly
statistically significant.

I'm sure
this isn't Ford's fault; he has the data and is probably far more aware
of its analytic pitfalls than we are.  But editors — Editors! —
have to begin thinking of stories as having many fascets, and work to
deliver the richest amount of data as possible that is related to the
stories and their context.


Course in crunching that health data
Sep 27th, 2005 by JTJ

Profs. David Kleinbaum and Nancy Barker will present their
online short course “Analysis of Epidemiologic Data” Oct.
14 – Nov. 11 at  Topics covered in the
course include: simple analysis of 2×2 tables, control of
extraneous variables (including an introduction to logistic
regression), stratified analysis, and matching.

David Kleinbaum, a professor at Emory University's Rollins
School of Public Health, is internationally known for his
textbooks in statistical and epidemiologic methods and as
an outstanding teacher.  He is the author of “ActiveEpi”
and “Epidemiologic Research- Principles and Quantitative
Methods” and has also taught over 150 short courses over
the past 30 years throughout the world.

Nancy Barker is a consulting biostatistician and a co-
author of the “ActivEpi Companion Text,” and has over 10
years of experience teaching short courses in epidemiology
and biostatistics at Emory University and the Centers for
Disease Control and Prevention.

As with all online courses at, there are no
set hours when you must be online, and you can interact
with the instructor over a period of 4 weeks via a private
discussion board.  We estimate you will need about 10 hours
per week.

Registration: $399 ($299 academic)


P.S.  Also coming up – “Clinical Trial Design” Oct. 21 –
Nov. 18 with Dr Vance Berger.
612 N. Jackson St.
Arlington, VA 22201

The basics of the basics: What is/are the definitions?
Aug 19th, 2005 by JTJ

Ford Fessenden, of the NYTimes, has yet another strong piece in Thursday's paper, “Health Mystery in New York: Heart Disease.”  The lede lays out the perplexing problem in NYC: “Death rates from heart disease in New York City and its suburbs are
among the highest recorded in the country, and no one quite knows why.”

But among possible answers — and here especially is where the AJ kicks in — is that there is some “…speculation that doctors in the area may lump deaths with more subtle
causes into the heart disease category, making that toll look worse
than it actually is
.”  And “…the Centers for Disease Control and Prevention, at the health
department's request, has sent specialists to determine whether doctors
in New York City ascribe causes of death substantially differently.”

I know, I know, we're preaching here, but we don't think it can be pointed out too often: journalists and all social scientists cannot simply accept given numbers as a true, valid, honest.  We always have to swim up the data-creation stream to determine where, why and from who came the numbers. 

Maybe "Performance Measurement" Isn't the Answer? At least if you are the one being measured.
Aug 2nd, 2005 by JTJ

We recently enjoyed meeting Stuart Kasdin at a Netlogo workshop
Stuart spent some years in the Peace Corps, then a decade with the OMB
(Office of Budget Management).  Currently he's working on his
doctorate in Poly Sci at UC-Santa Barbara.

Stuart has also been thinking about “performance measurement,” the
term-of-art used by auditors and managers of government agencies. 
(In the private sector, the term often used is “forensic
accounting.”)  We have generally thought well of performance
measurement, especially as a vocabulary and tool journalists should
know about to better understand and evalutate the performance of
government.  Stuart, however, has thought about this in greater
depth, and from the perspective of someone inside the government. 
His paper, “When Do Results Matter?  Using Budget Systems to
Enhance Program Performance and Agency Management” is worthwhile

: “Managing by results” is a widely used public
budgeting approach based on developing performance measures that display the
progress of a program toward its stated objectives.  This paper considers the complex environment of government
budgeting and how to establish budget systems that can successfully encourage
improved performance by managers.  The
paper assesses the limitations in how governments currently apply performance
budgeting and suggests ways that it might be made more effective.  First, performance measures must be individually
tractable and simple, as well as be coherent and revealing in the context of
other program performance targets. 
In addition, performance budgeting must distinguish between
program needs based on environmental changes and those based on management
related decisions.  Finally,
paper argues that
multi-task, complex-goal programs
will typically result in low-powered incentives for program managers.  This outcome results because, even apart
from information obstacles, program managers will be rewarded or punished on only
a component of the program, representing a small fraction of the total program
performance when performance measures as increase.  A partial solution is to ensure that the number of policy
instruments is not smaller than the number of targets.”   

Click here to read the Kasdin paper.

What we can learn from bioinformatics
Jul 10th, 2005 by Tom Johnson

A recent profile of mathematician-turned-geneticist Philip Green is a good-read introduction to bio-informatics, and bio-informatics just might produce some methodologies journalists can use to validate public records databases.

The article, “Bioinformatics,” is in the quarterly published by the Howard Hughes Medical Institute.  Some highlights:

* Using a detailed computational model, [researchers] found
that some kinds of [genetic] mutations occur at constant rates, like the
ticking of a clock, which makes them useful for dating evolutionary events.
Other kinds of mutations occur at varying rates de-pending on the generation
times of the organism. This information in turn makes it much easier to identify
parts of the genome that exhibit different patterns of change over time,
indicating that the DNA in those regions is subject to selection and therefore
playing a functional role. The idea, says Green, is to separate the noise of
meaningless changes in DNA so that the signals of consequential changes emerge
clearly from the background
Journalists could look at which elements are changed in a data base and
how often as a clue for the importance of the data base and the
relative importance of various elements.

* “The main issue [in biology and genomics] is how
quantitative we’re going to be able to get,” [Green] says. “Most people will
accept the idea that we will know qualitatively how things are interacting with
each other. But what you really want is a quantitative result, so that you can
change the levels of one component and predict how it will affect the system.”

*  “Back then, [says a colleague of Green’s] we wondered if
there was a need for mathematics in biology. In the mid-1980s, there weren’t a
lot of data. Biology was about analyzing the notes in your lab book.

the last 20 years, biology has become dominated by huge data sets. Now it’s an
exception rather than the rule to publish a paper that does not draw on large
databases of biological information. Mathematical analysis has become a
funda-mental part of biological research. It has turned out to be of equal
importance to experimentation.”

Take a look at the article.  It suggests some parallels of investigation for analytic journalism.

What's behind the curtain? "Private Warriors"
Jul 7th, 2005 by JTJ

We're pleased that the PBS program “Frontline” is keeping up the good fight to produce important journalism.  And thanks to the Librarian's Index to the Internet for pointing us to:

Private Warriors

This Public Broadcasting Service (PBS) Frontline program looks “at
private contractors servicing U.S. military supply lines, running U.S.
military bases, and protecting U.S. diplomats and generals” in Kuwait
and Iraq. Website features discussions of the appropriateness of
outsourcing, whether privatization saves taxpayer money, and the role
of contractors. Includes contractor profiles, interviews, a FAQ, video
of the program, and related links.

Subjects: Government contractors — United States | Public contracts — United States | Private security services | United States — Armed Forces — Management | New this week

Created by
je – last updated Jul 6, 2005

Be sure to drill down to the section, “Does Privatization Save Money.”  A nice example of a reporter asking the right questions.

Web Interface for Statistics Education (WISE)
May 26th, 2005 by JTJ

From the good ol' Librarians'
Index to the Internet
comes a good site/toolbox for learning and teaching

“The Claremont Colleges' “Web
Interface for Statistics Education” (WISE)
seeks to expand
teaching resources offered through Introductory Statistics courses, especially
in the social sciences. This project aims to develop an on-line teaching tool
to take advantage of the unique hypertextual and presentational benefits of the
World Wide Web (WWW). This teaching tool's primary application is as a
supplement to traditional teaching materials, addressing specific topics that
instructors have difficulty in presenting using traditional classroom
technologies. The tool serves to promote self-paced learning and to provide a
means for advanced students to review concepts.”

Doing well by doing good
May 19th, 2005 by JTJ

at the IAJ we believe one of the reasons people come to newspapers or
broadcast stations is to get the data which, upon analysis, they can
turn into information that helps them make decisions.  Ergo, the
more meaningful data a journalistic institution can provide, the
greater value that institution has for a community.

A good example arrived today thanks to Tara Calishain, creator of ResearchBuzz.  She writes:

** Getcher Cheap Gas Prices on Google Maps


when I was saying that I would love a Gasbuddy / Google Maps mashups
that showed cheap gas prices along a trip route?   Turns out
somebody has already done it —  well, sorta. You can specify a
state, city  (only selected cities are available) and 
whether you're looking for regular or diesel  fuel. Check it out

The data driving the map is ginned up by 
It's not clear how or why GasBuddy gets its data, but it offers some
story potential for journalists and data for news researchers.  It
has an interesting link to dynamic graphs of gas prices over time.

Surely the promotion department of some news organization could grab
onto this tool, tweak it a bit,  promote the hell out of it, and
drive some traffic to and build loyalty for the organization's web

That's the obvious angle, but what if some enterprising journo started
to ask some questions of the data underlying the map?  What's the
range in gas prices in our town/state?  (In Albuquerque today, the
range was from $2.04 to $2.28.)  Are there any demographic or
traffic flow match-ups to that price range?  How 'bout the
variance by brand? 

Would readers appreciate this sort of data?  We think so,
especially if there was an online sign-up and the news provider would
deliver the changing price info via e-mail or IM much like Travelocity
tells us when airline ticket prices change by TK dollars.

»  Substance:WordPress   »  Style:Ahren Ahimsa