Chance News 53
From Bayesian statistics for experimental scientists written by Hal Stern, Department of Statistics, University of California, Irvine:
If they would only do as he did and publish posthumously
we should all be saved a lot of trouble
in reference to followers of Rev. T. Bayes
If your experiment needs Bayesian statistics,
you ought to have done a better experiment
N. Gilbert (Biometrical Interpretation, 1973)
Submitted by Paul Alper
From a Wall Street Journal article  describing the results of a number of research studies about victims of investment fraud:
The typical investment-scam victim is an optimistic married man in his later 50s who has a higher-than-average knowledge of financial matters and deep confidence in his own judgment ….
[A man told] an FTC fraud forum that he preferred speaking with a man because "you can lather him up and push all the green buttons." Women were more cautious and asked too many questions, he said, prompting an office maxim, "Don't pitch to the b—."
NYT headline: "For Todays Graduate, Just One Word: Statistics"
You can read the article Here.
The article is about the Joint Statistical Meetings in Washington this week and how the job market for statisticians is booming. This is said to be a product of the recent explosion of digital data. Here is the accompanying picture:
A brief Wall Street Journal article  summarized a JAMA report about two recent studies of the relationship between the Mediterranean Diet and cognitive decline. Apparently an (undated) older study had suggested the existence of a relationship.
One of the two recent studies involved 2,000 elderly people and found that “those who adhered more closely to a Mediterranean diet … had less risk of developing [dementia].” The other study, from France, involved 1,410 people and found that “[a]dherence to the diet didn’t change the risk of dementia.”
JAMA’s editorial conclusion:
[A]ll told, there is “moderately compelling evidence that adherence to the Mediterranean-type diet is linked to less late-life cognitive impairment.”
Kuklo's Fellow Infuse Worker
From The Pioneer Press we learn that there is more to the Kuklo story. "Dr. David Polly, the University of Minnesota spine surgeon ... received nearly $1.2 million in consulting fees from medical device giant Medtronic over a five-year period." The details "of Polly's billing records were released this week by Sen. Charles Grassley, R-Iowa, as an attachment to a letter to University of Minnesota President Robert Bruininks. The letter raised questions about how the U polices conflicts of interest among doctors."
Polly's recordkeeping was indeed detailed:
Download CDs from meeting, 15 minutes, $125
Dinner meeting, 240 minutes, $2,000
E-mail Medtronic employee, five minutes, $49.48
Conference call, 90 minutes, $890.63
Teach at scoliosis meeting, 330 minutes, $2,750
According to the newspaper, Dr. Charles Rosen, a spine surgeon in California who leads a medical ethics group, said he was among those surprised by the details.
"I've not seen anybody bill the way he did," said Rosen, of the University of California-Irvine, who acknowledged that he doesn't do paid consulting work with the device industry.
"In my opinion, it sounds more like an investment banker," he said of the detailed billing. "It doesn't sound like someone in medicine."
Submitted by Paul Alper
Defining a clunker
“When Precision Is Only 92.11567% Accurate”
by Charles Forelle, The Wall Street Journal, August 5, 2009
Temporarily substituting for Carl Bialik (The Numbers Guy), Forelle reports about the government’s cash-for-clunkers program and critiques the EPA’s recently revised definition of a clunker.
The EPA stated that “more precise” data calculated “to four decimal places” led it to revise its miles-per-gallon cutoff figures.
"It is ludicrous to suggest that you can get fuel-consumption accuracy anywhere past the first decimal place, let alone the second," says … an independent U.K. auto tester.
Forelle discusses the “faux precision” of estimates that are often based on sampling, but reported as final counts or measurements without their sometimes large margins of error, as in the case of population or unemployment-rate estimates.
He cites another issue involved in misleadingly precise estimates, that is, lack of adherence to conventional rules relating to the issue of significant figures in arithmetic.
The principle is simple: When combining measured numbers, the final answer is only as precise the least-precise piece of data that went into it; you can't just add a tail of decimal places, even if they show up on the calculator. So a room that's 2.5 meters (two significant digits) by 3.87 meters (three) has an area of 9.7 square meters, though the two numbers multiply to 9.675.
Fuel mileage calculations are apparently based upon tailpipe emissions of carbon dioxide, because released carbon dioxide from burning fuel provides a more accurate measure of gas consumption than direct measurement of consumed fuel. Not only does the EPA believe that the results of two lab tests on each car must be recorded to four decimal places by law, but it also added tests that were not done on older cars, and “created a formula that estimated from the old data what would happen had the new tests been run,” this despite the different precision levels of numbers that went into the formulas. An EPA spokesman said, "Repeatability and accuracy is something we spend a lot of time on."
1. What’s the difference between accuracy and precision?
3. A blogger commented , “Regards your room area example, if both length and width are measured by a person who makes the same direction of error on each measurement -- so that both are either too high or too low -- then the area will not only have almost twice the percentage error of either measurement, it will, on average be too high.” Do you agree with all, or part of, this statement?
4. The author described a 1991 court case in which an Alaskan man failed a bar exam and “missed by 0.5 point the threshold needed for a re-evaluation of his test.” The man claimed that, since the essays were graded with integers, his score should have been rounded up to the next integer. Although the man lost the case, the Alaska Supreme Court found his argument “convincing from a purely mathematical standpoint.” A blogger argued  that there are an infinite number of significant digits in counts (e.g., 1.000…), because “the error in the value of these numbers is ZERO,” and so arithmetic results “can be rounded to as many sig figures as you want to.” Do you agree with the Alaskan man or with the blogger?
All that jazz?
“Can Jazz Be Saved?”
by Terry Teachout, The Wall Street Journal, August 9, 2009
Based on the National Endowment of the Arts’ 2008 Survey of Public Participation in the Arts , popular interest in jazz on the part of adult Americans appears to be experiencing a serious decline. The study was conducted “in participation with” the U.S. Census Bureau.
Several causes for concern about the future of jazz are a general decrease in attendance in at least one jazz performance per year (down from about 11% to 8% for the period 2002-2008) and an increase in the median age of those who do attend (up from 29 to 46 years for the period 1982-2008).
Supplementary materials  include a brochure, trend tables (1982-2008), survey instrument, data user’s guide with information on the survey design and procedures, and raw data file.
Irresponsible data mining
“Data Mining Isn’t a Good Bet For Stock-Market Predictions”
by Jason Zweig, The Wall Street Journal, August 8, 2009
Columnist Jason Zweig discusses “quantitative money manager” David Leinweber’s new book, Nerds on Wall Street: Math, Machines and Wired Markets  (Wiley, June 2009).
In his book, Leinweber “dissects the shoddy thinking that underlies most of [the quantitative] techniques” in use today, and he refers to data-mined numbers as “one of the leading causes of the evaporation of money.”
Zweig describes how Leinweber decided to satirize data mining with an example, meant to be a joke. He found that annual butter production in Bangladesh “explained” 75% of the variation in the annual returns of S&P 500-stock index over a 13-year period.
By tossing in U.S. cheese production and the total population of sheep in both Bangladesh and the U.S., Mr. Leinweber was able to "predict" past U.S. stock returns with 99% accuracy.
Leinweber has advice for avoiding “falling into a data mine”: (a) Check that the results make sense; (b) Check that the claim still holds for smaller subsets of the data; (c) Check the results after costs, fees, and taxes are subtracted; (d) Wait to see if the claim continues to hold true as time goes by.
Students may enjoy a 4-minute video  of Jason Zweig interviewing Nerds author David Leinweber. Or they may be interested in YouTube videos (8 minutes, in four parts)  ] of a lecture by David Leinweber.
The article contained a chart  showing the “Correlation of Super Bowl wins by original NFL teams with positive return[s] for the S&P 500.” The bar chart shows the S&P 500 return for each year 1967 through August 6, 2009, with 32 blue bars for years in which a “correlation held” and 11 red bars for years in which it did not.
1. What do you think the chart’s author meant by stating that a “correlation held”? What else would you like to know about his/her “correlation”?
2. Suppose there were a positive correlation, even a relatively high one, between NFL Super Bowl team wins and positive S&P 500 returns. Would you be surprised that Super Bowl wins did not predict positive S&P returns for every one of the years?
Football: How about a “time-in”?
“Start the clock: A modest proposal for improving football: the 'time-in'”
by Samuel Arbesman, Boston Globe, August 16, 2009
The author, a postdoctoral fellow at Harvard Medical School, proposes a new football rule - the “time-in” - which would force the game clock to resume. He would limit its use to once per game per team.
The possibility of a sudden time-in would loom large in every coach’s mind at the most tense points in the game, introducing just enough concern and uncertainty to make the game different.
On his website Advanced NFL Stats , Brian Burke has developed a mathematical model for computing the probability of victory for a team during a game. According to Burke, the time-in would “only be used by the leading team on defense near the end of the game when there’s a small point difference.” In that case, his prediction is that the time-in “could produce up to a one-third drop in win probability for the losing team.”
On the other hand, if the time-in has not yet been used during a game, it could be a potential threat to an opposing team, who might have to face the “Unexpected Hanging paradox”:
Imagine a prisoner is told that sometime during the next week he will be hanged, and it will be a complete surprise. The prisoner … reasons thus: I can’t be hanged on Friday, because it’s the final day of the week, and therefore not unexpected. So, I can only be hanged sometime between Monday and Thursday. However, it can’t be Thursday, because now that’s the last possible day to be hanged, and so it won’t be a surprise then either. Continuing this train of thought, the prisoner coolly deduces that he can’t be hanged any day of the week at all, and therefore will not die. He is therefore quite surprised when he is woken up early on Wednesday and sent to his death.
The author feels that the surprise element of a time-in would provide more excitement for fans.
He cites John Reed’s Clock Management as “the authoritative ... book on time in football.”
“Your Baby Is Smarter Than You Think”
by Alison Gopnik, The New York Times, August 15, 2009
“Babies’ brains are more sophisticated than we ever believed”
by Alison Gopnik, [UK]Times Online, August 15, 2009
The author is a professor of psychology at Berkeley and author of The Philosophical Baby.
She describes three recent experiments that show that “even the youngest children have sophisticated and powerful learning abilities” and “show how brilliant a baby’s mind really is.”
In 2008 researchers at the University of British Columbia “proved that babies could understand probabilities.” One-year-old babies were shown a box that contained red balls (20%) and white balls (80%). The babies appeared more surprised (“looked longer and more intently at the experimenter”) when an experimenter pulled a sample of mostly red balls (80%) from the box of mostly white balls.
The babies concluded that the researcher must like the red balls more than the white ones as when she held out her hand, they gave her a red ball rather than a white one. Far from being illogical and egocentric they could learn from statistics and use the logic of what they saw to figure out what someone else wanted.
In 2007 researchers at M.I.T. “demonstrated that when young children play, they are also exploring cause and effect.”
One group [of pre-schoolers] was shown that when you pressed one lever, the duck appeared and when you pressed the other, the puppet popped up. The second group observed that when you pressed both levers at once, both objects popped up, but they never got a chance to see what the levers did separately, which left mysterious the causal relation between the levers and the pop-up objects.
When the children were subsequently given the toys to play with, the latter group were more interested in playing with it and “just by playing around, they figured out how it worked.”
In 2007 the author, at Berkeley, “discovered that pre-schoolers can use probabilities to learn how things work.” The children were shown two blocks that could light up a machine, with a yellow one making the machine light up 2 out of 3 times and a blue block making it light up only 2 out of 6 times. When the children were given the blocks and asked to light up the machine, they were more likely to chose the yellow block.
These astonishing capacities for statistical reasoning, experimental discovery and probabilistic logic allow babies to rapidly learn all about the particular objects and people surrounding them.
The author also advocates letting babies experiment through play, developing their “capacity for statistical reasoning, experimental discovery and logic,” rather than enrolling them in programs and buying products that claim to “make their babies even smarter.”
A blogger  commented:
In the experiment you mention with the red and white pinpong balls. It would be interesting to know if the results were the same when the colors were switched, or was it simply a question of the color red?