!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CHANCE News Chance news 2.15
(20 August to 15 September 1993)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Prepared by J. Laurie Snell as part of the CHANCE Course
Project supported by NECUSE and the National Science
Foundation.
Please send suggestions to dart.chance@dartmouth.edu
Current and previous issues of chance news and full text of the
newspaper articles can be found on the chance gopher. Just point
your gopher to chance.dartmouth.edu
===================================================
Where observation is concerned, chance favours only the
prepared mind.
Louis Pasteur 1822
===================================================
ARTICLES ABSTRACTED
>>>>>==========>> This quote and the following article were suggested by Goran Djuknic. I appreciate getting suggestions. How to live forever. Esquire Magazine, Sept 1993, pp 125-132 Michael Segell The author uses results of recent studies to give advice to men on how to live longer. It is a two part series and the second part will appear in the November Esquire. This part itself has three parts. The first part deals with behavioral suggestions: stay married, be an optimist, make a lot of money etc. suggested by studies like the Harvard Study. The second part compares sedentary men with active men at ages 32 and 65 with respect to metabolism, cholesterol, blood pressure, etc. The third part concentrates on the health of your heart. Not a bad summary if you can stand the Esquirese writing. <<<========<<
>>>>>==========>> The next item appeared on Edstat sent by Bret Presnell. Pollsters enlist psychologists in quest for unbiased results. The New York Times, 7 Sept 1993, C1 Daniel Goleman This article talks about various sources of "bias" (never defined, of course), and particularly about wording of questions. It gives some pretty good examples which you might be able to use in class discussion. Of course, they get at least one thing wrong, but that's even better for discussion. Specifically, an accompanying article, "Experts Advise a Healthy Skepticism About Polls," states that "if the margin of error is three percentage points in a pre- election poll, that means the number of people who support a given candidate might be three percentage points higher or lower than the result in the poll." Nothing about the confidence level even, much less what it means. This is in fact very surprising, since the New York Times is usually quite careful about explaining this kind of thing when they report their own (NYT/CBS) poll results, even explaining that "in 19 samples out of 20, ..." <<<========<<
>>>>>==========>> Larry Lester's upcoming course at the University of Texas on lotteries got national attention in the news when the Associated Press picked up from the following article: Professor seeks to even odds of lottery: Class aims to unravel the numerical mysteries of the state's game of chance. Austin American-Statesman, 28 Aug 1993, City/State p 1 A description of an informal not-for-credit course called "Lotto Luck" to be offered next term at the University of Texas. The professor, Larry Lesser, will discuss the statistical, psychological, and marketing problems involved in gambling with emphasis on lotteries using the Texas Lottery as a basic example. He says "I want people to make an informed choice before deciding whether to play the Texas Lottery or other games of chance, learn the underlying principles of probability. I'm pro-choice not necessarily pro-chance." The article contains Lesser's answers to a number of common questions asked about lotteries. <<<========<<
>>>>>==========>> The next three items were provided by Bill Peterson Ask Marilyn. Parade Magazine, 25 July 1993, p 18. Marilyn vos Savant Two more statistical questions for Marilyn, neither as controversial as some of her earlier postings, but interesting nonetheless. First, a reader had questions about an illustration, reproduced in the column, on consumption of alcoholic beverages by men and women. The graphic, using bottles to create stacked bar charts, shows men consuming 84% of all beer (women 16%), 54% of all liquor (women 46%), 49% of all wine (women 51%), and 41% of all wine cooler (women 59%). The reader was puzzled by a caption that read: "Men consume about 76% of all alcoholic beverages." The reader proposes that this might mean (1) of people who drink, 76% are men and 24% are women, or perhaps might be corrected to (2) of all alcohol consumed 57% is by men (averaging 84%, 54%, 49% and 41%). Marilyn correctly points out both of these would be wrong. The first misinterprets what is being measured, the second fails to account for the fact that the four types are not sold in equal amounts, noting that if more beer and liquor are sold than wine beverages, the 76% cited can be correct. The second question is a standard-coin tossing problem. The reader recalls that Marilyn once said the chance of tossing heads after observing 5 heads in a row was still one out of two. But she relates a story about a Super Bowl announcer who flipped a coin that came up heads and then noted that he had flipped the coin five times before the game, getting heads each time. He said the chances of this happening are one in 68. Marilyn accurately distinguishes between the chances before the tossing starts, and the chances after the 5 heads are observed (and also corrects the above to one in 64). <<<========<<
>>>>>==========>> Bottle cap flap riles the masses. Los Angeles Times, 26 July 1993, A1 Bob Drogin Not a probability story proper, but an interesting ethical dilemma with promotional marketing games. Last year Pepsi ran an ad campaign in the Philippines based on a 1 million peso prize (about $40,000 U.S.) for the holder of a specially marked bottle cap. When the lucky cap number 349 was announced, winners began flooding in. In what has been called the greatest marketing mistake ever, it seems that 800,000 caps numbered 349 were printed, and more than 486,000 claimants have already appeared. Rioting ensued when the error was announced, and panicked executives hastily offered to pay $20 in pesos to each "winner." Not surprisingly, this offer has done little to soothe the populace. The article reports that criminal and civil claims filed against Pepsi outnumber those against the late President Ferdinand Marcos, who is accused of looting the nation of $10 billion while in power. (But observe that some quick arithmetic puts Pepsi's prize liability here at $32 billion). <<<========<<
>>>>>==========>> A reader's response. Science 261 (30 July 1993), p 533. Letters Katherine H. Lowe An exchange between a reader and an editor over Science's reporting of a reader survey related to the special report on women in science. An earlier editorial had argued for the value of such a survey as a "gauge of the opinions of [Science's] most committed readers." Nonsense, says this letter to the editor: only proper statistical survey sampling will produce any useful information. In his rejoinder the editor says "information of any sort can be useful as long as it does not pretend that it is more extensive or important than its intrinsic worth." He adds that statisticians did not invent the word "survey" and have no claim as unique arbiters of its usage. This could be an interesting topic for class discussion. Incidentally, preliminary results from the survey which asked questions about women and science can be found in an editorial note at the end of Letters in the 23 July issue of Science. <<<========<<
>>>>>==========>> And then the patients suddenly started dying. Washington Post, 7 Sept. 1993, A1 John Schwartz This is an account of a recent drug trial that ended in a catastrophe. The drug being tested was Fialuridine (FIAU) that companies have been trying to find a use for since 1970. As a last try it was tried out on 43 patients who had both the HIV virus and Hepatitis B. There was a significant reduction in hepatitis and this led researchers to additional trials. Approval was given by the FDA in March 1992 to give the drug to 24 hepatitis B patients at four doses. Patients with low doses did not see much advantage and those with high doses tended to get nauseated while those in the middle made remarkable progress, so a longer trial was planned. Even during the planning stage for this second trial, some problems arose in the previous study that, in hindsight, should have caused more concern than they did. However, the new study went ahead this Spring and by late June it was clear that it was a disaster and was stopped. By the end of August five participants had died. A chemical like FIAU is introduced to try to take over the mechanism by which the offending virus duplicates its DNA. This approach has had some success in the use of AZT to combat the HIV virus. It was hoped that the FIAU itself would not cause damage to other parts of the body. In this case the researchers were wrong and the drug destroyed livers, pancreas and other major organs in the process. The article provides a moving account of involvement of patients, the director of the project, and other scientists in realizing that things had gone very wrong and in finding out what went wrong. A congressional hearing is planned to see if proper procedures were followed. <<<========<<
>>>>>==========>> In medical research, physicians' loyalties sometimes conflict. The Washington Post, 7 Sept. 1993, A8 David Brown Obviously inspired by the incident described in the previous article, this article reviews some of the ethical issues involved in clinical trials. For example, some medical ethicists feel that the doctor or researcher involved must be convinced that a would-be subject is as likely to benefit from the conventional treatment (or placebo) as from the new one. This is called "equipoise." Others say that randomized trials are ethical as long as there is an honest uncertainty in the medical community as a whole as to what treatment is best for the illness involved. Individual investigators do not have to have this perfectly balanced state of mind. Another issue is whether an investigator should also be allowed to be a personal physician of a subject while the trial is going on. <<<========<<
>>>>>==========>> Here is a related but more technical article; Implications for trials in progress of publication of positive results. Lancet, 11 Sept 1993, pp 653-657 Fletcher and others The authors address the following questions. Several major trials are being carried out more or less simultaneously to test a new drug or procedure. One study announces a positive results. What effect does this have on the other ongoing experiments. In particular, how positive a result is necessary for the other studies to feel compelled to stop their study or at least inform the subjects that the original uncertainty about the drug is no longer valid. The authors discuss an actual example where this has happened and where some of them are on the data-monitoring committee for other studies that have to make this decision. They illustrate a bayesian approach to trial- termination decisions. <<<========<<
>>>>>==========>> DNA tests fizzle; murder case may fold. Sacramento Bee, 16 Sept, B1 Ramon Coronado The Sacramento County District Attorney's Office asked a judge to dismiss charges against John Bertsch and Jeffrey Hronis in a 1985 kidnap, rape and murder case. The Deputy District Attorney learned the FBI itself was not in agreement as to what the appropriate estimates are for the probability of a match in this case. Its DNA tests in 1989 showed that the chances of a match were 1 in 12 million for Bertsch and 1 in 8 million for Hronis. In a retesting in 1992 the FBI came up with 1 in 16000 for Hronis and 1 in 200 for Bertsch. Hayes made her announcement to drop the charges at what was planned to be a hearing on the merits of DNA fingerprinting. The judge granted a continuance for one week to consider the defense request that the prosecutor hand over the FBI's DNA test results. Hayes claims that if the case is dismissed the defense is not entitled to these results. The defense is worried that it is all a ploy to delay a trial until the issue of use of DNA fingerprinting gets clarified. Stay tuned. <<<========<<
>>>>>==========>> Gene clues offer hope of a DNA 'photofit'. The Independent, 1 Sept 1993, Home news, p 4 Susan Watts At the annual meeting of the British Association for the Advancement of Science, Alex Jeffreys, the inventor of DNA fingerprinting, predicted that, in the near future, DNA fingerprinting will be able to reveal the owner's physical characteristics. At the moment, sex can be determined easily but not much else. It may soon be possible to determine ethnic group, but, Jeffreys said "That is a form of pigeon-holding many will find deeply repugnant". He also warned that looking for something like the "gay gene" would bring the whole field into disrepute. <<<========<<
>>>>>==========>> Conversations/ David Card and Alan Krueger. The New York Times, 22 Aug, 1993, Sec 4 Page 7 Sylvia Nasar Two Princeton economists, David Card and Alan Krueger, have decided to employ classical statistical methods using control groups etc. to test some standard economic theories. For example, they tested the theory that a higher minimum wage increases the pay of some low-wage workers but forces others out of work, let off by companies that cannot afford to pay them. They tested this when New Jersey, in the middle of a recession, raised the minimum wage in the state making it 80 cents higher than in the neighboring state Pennsylvania. They surveyed 400 fast-food restaurants and found that the New Jersey restaurants, on average, added 2.5 workers after the minimum wage went up, while in Pennsylvania the payrolls decreased. They suggested the explanation that when the minimum wage is too low, restaurants typically have a couple of positions they can't fill, but when the minimum wage goes up "teens who are sitting home go out looking for jobs." The article claims that the use of test of hypotheses and such things are new to theoretical economists and that the above study among others has had some influence on the actions of the Clinton administration. <<<========<<
>>>>>==========>> More about doom. The Mathematical Intelligence, Summer 1993 Letters to the Editor Evidently the Gott craze to estimate the time of the end of the world was started by Brandon Carter, an Applied Mathematician at Cambridge. Carter never actually published exactly what he had in mind but gave talks about it and John Leslie at Oxford has described what he thinks Carter had in mind. Carter and Leslie argue that we over-estimate how long we are going to be around and suggest the following simple Bayesian argument to show why they believe this. Assume that we will last either a relatively short time, like to 2150 or some much longer time thousands of years later. If we think of ourselves as at a random time in the lifetime of our civilization then our prior probability places high probability on the longer time. But then we calculate the posterior probabilities, taking into account that we are here in 1993 and we find that the shorter time has the high probability showing that our prior estimate was too optimistic. Dieks criticizes this argument claiming that it uses the information that you are here now in 1993 in setting up the priori probabilities since you chose only times after 1993. Therefore you can't use this information again to get new priori probabilities. Dieks provides an argument that he thinks gets around this problem. He assumes first you do not even know if you are human and so assign priori probabilities fifty fifty for the two outcomes. You then learn you are human and reason that you are much more likely to be in the longer period, so you change to the more optimistic assignment of probabilities. Finally, you learn that it is 1993 and mankind has not been around a long time so you end up with your final posterior probabilities back to the more pessimistic probabilities. The letters contain additional references. The issue raised by Leslie of the proper use of information in Bayesian analysis seems interesting. <<<========<<
>>>>>==========>> Women and alcohol: heart benefit versus cancer risks. The New York Times, 15 Sept, C13 Jane E. Brody The recent report from the Nurses' Health Study of 89,000 middle aged women found that women who consumed three to nine drinks a week were 40 percent less likely to develop heart disease than non-drinkers. Another ten year study of about 130,000 men and women found that those who had one or two drinks a day had a 30 percent less chance of dying from heart disease than did non-drinkers. On the other hand, again from the Nurses' Health Study, women who had three to nine drinks a week were 30 percent more likely to develop breast cancer than were non-drinkers. Where does that leave a women, trying to decide if she should drink or not? The experts consulted pointed out that, from the ages of 50 to 70, the coronary death rate in women is two to four times the death rate from breast cancer. In addition, one study suggested it was the drinking before 30 that accounted for most of the breast cancer danger. Thus, except for the obvious warnings not to drink when pregnant, not to mix with drugs etc., the article seems to suggest that women over the age of 30 who enjoy moderate drinking might as well continue enjoying their drinks. <<<========<<
>>>>>==========>> Vitamins cut cancer deaths in large study in rural China. The New York Times, 15 Sept 1993, C13 Warren E. Leary A report being published in the current Journal of the National Cancer Institute describes the result of a cooperative study between American and Chinese researchers on the use of vitamins to prevent cancer. The subjects were 29,584 Chinese ages 40 to 69 who are residents of Linxian County in Henan Province. The subjects were randomly assigned to eight groups, with seven groups receiving different combinations of vitamins and minerals in doses one to two times higher than the recommended daily doses in the United States. The eighth group received placebo. The group with the highest benefits got the beta carotene supplements that include vitamin A, vitamin E and the mineral selenium. This group had a 13 percent drop in cancer death rate and 10 percent less deaths from strokes. Researchers said that this was the first randomized trial to show a significant reduction in cancer through vitamin supplements. However, they noted that the reduction occurred in a population that had an unusually high cancer rate and nutritionally deficiency. Thus, the results might not be so significant for those in the United States who have diets much richer in healthful nutrients and vitamins and suffer from different kinds of cancer. Similar studies are underway in the United States but conclusive results are years away. <<<========<<
>>>>>==========>> 'Perspective on education in America': Sandia study challenges misconceptions. Skeptical Inquirer, Fall 1993 Kendrick Frazier The study "Perspectives on Education in America" was carried out by members of Sandia National Laboratory. It put a brighter light on American education than other studies have, and some have suggested that the previous administration held up its publication because they wanted, for some reason, to present the public schools system in a bad light. This article discusses the finding that, from 1975 to 1991, the average SAT scores were declining, but the scores of whites remained stable and the minority sub populations showed improvements -- a kind of Simpson's paradox. The study showed that, when SAT scores are controlled for such things as class rank and gender, the average performance improved 30 points during this period. <<<========<<
>>>>>==========>> Diagnoses of alien kidnappings that result from conjunction effects memory. Skeptical Inquirer, Fall 1993 Robyn M. Dawes and Matthew Mulford Tversky and Kahneman have noted that the conjunction of two events is often viewed to be more probable than one of them alone. The point of this note is that this can play a role in surveys. The following question appeared on a Roper study. How often has this happened to you: Waking up paralyzed with a sense of a strange person or presence or something else in the room? The authors carried out a survey where 144 subjects were asked this question and another 144 subjects were asked simply How often has this happened to you: Waking up paralyzed? For the conjunction of "Waking up paralyzed and with a sense of a strange person etc." 40 percent answered that this had happened at least once. For "Waking up paralyzed" just 14 percent said that this had happened at least once. Apparently the Roper question has been cited by other researchers as evidence of post-traumatic stress from kidnapping by aliens. <<<========<<
>>>>>==========>> 'Power-Line Cover-Up': high-voltage charges, weak case. Hartford Courant, 9 Sept 1993 Joel Lang This is a review of Paul Brodeur's book "The great power-line cover up: how the utilities and the government are trying to hide the cancer hazards posed by electromagnetic fields" published recently by Little Brown. In the last chance news I mentioned the book "Toxic Terror" by Wheelan and "Phantom Risk" edited by Foster, Bernstein and Huber. These two books together with Brodeur will give you the complete range of feelings on the issue of dangers from power lines. Whelan takes the position that the dangers are greatly exaggerated by the press etc., "Phantom Risk" provides a more technical assessment in terms of legal cases, and Brodeur goes off in the direction of cover up etc. Much of Brodeur's feeling about this issue have appeared in his articles in the New Yorker. <<<========<<
>>>>>==========>> Test after colon surgery appears ineffective. The New York Times, 25 Aug. 1993 Associated Press The current issue of the Journal of the American Medical Association reports on a study showing that a blood test called CEA commonly used to detect new malignancies after surgery for colon cancer is not very effective. The researchers found that, while the test often warns that cancer has reappeared in the colon or elsewhere, it also misses many recurrences. In addition it suggests malignancies when none exist. Even when it is right, it is often too late and, when it is wrong far more expensive tests are necessary (and sometimes surgery) to rule out cancer. It had been hoped that CEA monitoring might save one in five patients with recurring cancer. But experts suggest that this study shows that, at best, it has very small effects at very large costs both in human terms and in dollars. <<<========<<
>>>>>==========>> Two studies report heart care lags for blacks. The New York Times 26 August, 1993, B8 Tim Hilchey Two reports in a recent issue of the New England Journal of Medicine compare the treatment and outcome of heart problems for blacks and whites. The first study was carried out at the Pittsburgh Department of Veterans Affairs Medical Center where it was found that, among veterans treated for heart ailments, white males were more likely to relieve advanced X-ray imaging, undergo balloon procedures and have bypass surgery. This was true even after adjusting for factors such as age, smoking etc. The Chicago study looked at 6,451 patients who suffered cardiac arrests outside a hospital in 1987 and 1988. Only about 2 percent of these survived, but whites were three times more likely to survive than blacks. Both reports carefully rule out many explanations other than simply discrimination but stop very slightly short of saying they believe it is the result of discrimination. <<<========<<
>>>>>==========>> A question of attribution. Lancet, 28 August, 1993 Fact, Figures, and Fallacies column J. R. Glynn Another in the ongoing series to educate doctors on basic epidemiological concepts. This one reviews the problem of determining when an association between an exposure and a disease is established whether there is a causal relation involved. As in previous discussions in the column the ideas are illustrated with lots of real examples from the medical literature. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! CHANCE News Chance news 2.15 (20 August to 15 September 1993) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Please send suggestions to: jlsnell@dartmouth.edu >>>==========>>|<<==========<<<
>>>==========>>|<<==========<<<