!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CHANCE News 3.04
(6 March to 20 March 1994)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Prepared by J. Laurie Snell as part of the CHANCE
Course Project supported by the National Science
Foundation and the New England Consortium for
Undergraduate Science Education.
Please send suggestions to:
dart.chance@dartmouth.edu
Current and previous issues of CHANCE News can be found
on our chance gopher. Just point your gopher to:
chance.dartmouth.edu
==========================================
You should treat as many patients as
possible with the new drugs while
they still have the power to heal.
Armand Trousseau
19 Century French physician
============================================
<<<========<<
>>>>>==========>> Note: Last time we had several articles from the journal "Teaching Statistics". This is an international journal for teachers of statistics published three times a year. Their address is: Teaching Statistics Department of Probability and Statistics University of Sheffield Sheffield S3 7RH, England This week we have a number of articles from "Chance Magazine" published quarterly by Springer-Verlag Publishers and the American Statistical Association Both these journals should be in your libraries! ========================================== <<<========<<
>>>>>==========>> Ask Marilyn. Parade Magazine, 20 March 1994, p31 Marilyn vos Savant
A reader wonders why the probability of rain is not always 50% since there are only two outcomes. Marilyn points out the absurdity of this, commenting that this kind of logic would lead to a 50% chance that the sun will not rise tomorrow. She goes on to say: But rain doesn't obey the laws of chance; instead, it obeys the laws of science. It would be far more accurate for a meteorologist to announce, "There's a 25% chance that a forecast of rain will be correct." DISCUSSION QUESTIONS: (a) What do you think Marilyn means? Do you agree with her? (b) How does your local weather forecaster decide on the probability for rain tomorrow? (Incidentally, if anyone REALLY knows how this is done I would love to hear the answer). <<<========<<
>>>>>==========>> High court reviews test to determine reasonable doubt. The New York Times, 23 March 1994, p. 1 Linda Greenhouse
Greenhouse reports that in two death penalty cases the Supreme Court has upheld the use of an instruction to juries which uses the phrase "moral certainty" in defining reasonable doubt. Nevertheless, at the same time the Court warned that this definition is outdated and possibly confusing. Writing in the majority opinion, Justice Sandra Day O'Connor said that the Court did not "condone" the use of the phrase "moral certainty", although its use was not unconstitutional in the cases under consideration. (In 1990 the Court found unconstitutional instructions to jurors which, in part, required of them "not an absolute or mathematical certainty, but a moral certainty.") O'Connor also said that the meaning of the phrase had changed over time and today "a jury might understand the phrase to mean something less than the very high level of probability required by the Constitution in criminal cases." In a concurring opinion, Justice Ruth Bader Ginsburg noted that several Federal appellate circuits have instructed judges not to define reasonable doubt at all, thereby avoiding the problem altogether. Ginsburg prefers instead finding a better definition, and endorses one proposed by the Federal Judicial Center, the research arm of the Federal Judiciary. This definition does not use the phrase "moral certainty", and says in part that "proof beyond reasonable doubt is proof that leaves you firmly convinced of the defendant's guilt. DISCUSSION QUESTIONS: (a) How would you define the phrases "moral certainty", "mathematical certainty", "reasonable doubt", "firmly convinced"? (b) Does it make sense to attach probabilities to these phrases, such as "moral certainty means at least a 99% chance of being true"? <<<========<<
>>>>>==========>> Two major stories in the newspapers the past two weeks dealt with the breast cancer studies. There both relate to a large multisite study on breast and bowel cancer, directed by Dr. Bernard Fisher of the University of Pittsburgh. The first story started with an article written by Pulitzer Prize winning reporter John Crewdson, of the Chicago Tribune. Crewdson started Gallo's problems by suggesting that he may not deserve the credit he got for the discovery of AIDS. This led to a four-year investigation of charges of scientific misconduct by the Office of Scientific Investigation which ended by finding no misconduct on the part of Gallo. <<<========<<
>>>>>==========>> Fraud in breast cancer study; Doctor lied on data for decade. Chicago Tribune, 13 March 1994 John Crewdson
A 1989 New England Journal of Medicine article on breast cancer therapy presented findings from a study directed by Fisher. Recurrence rates for cancer after lumpectomy and radiation were compared with rates following mastectomy. The study involved 5,000 patients at nearly 500 medical centers. The recurrence rates were found to be essentially the same, leading many women to opt for the less disfiguring lumpectomy and radiation treatment. This article reported that Federal investigators discovered that one of the principal collaborators, Dr. Roger Poisson at the University of Montreal, enrolled at least 100 of his cancer patients who were not eligible for the study and falsified their records to make them appear eligible. It is suggested that Poisson was not trying to affect the outcome but only wanted to appear good at recruiting patients, in order to improve his chances for continued financial support. Poisson contributed about 16% of the patients to the 1989 study. The organizers of the study privately assured investigators nearly two years ago that the fraud did not affect the results of any of their findings. Fisher promised to publish a re-analysis of the data. The article reports that when he was asked "how soon", he replied "We don't know yet". Later news articles reported that Federal health officials have commissioned three statisticians to analyze the study independently and to give their report in two weeks. Fisher is now expected to give his own analysis to the New England Journal of Medicine within the next two weeks. DISCUSSION QUESTION: Fisher's team stated that they re- analyzed the data excluding Poisson's data and found that the study's published conclusions remained valid. But a government spokesman remarked that the remaining sample diminished the statistical power of the study. What does all this mean? <<<========<<
>>>>>==========>> Flawed study raises questions on U.S. research. The New York Times, 15 March 1994, p. C14 Lawrence K. Altman
Altman remarks that there is a widespread misconception that scientific journals examine the raw research data. He says that the fact they do not makes it difficult to detect fabrications. He gives a detailed discussion of how the irregularities were found in the Poisson data and of Poisson's explanations for why he registered patients who were not eligible for the study. <<<========<<
>>>>>==========>> Data on risks creates debate about drug to prevent breast cancer. The New York Times, 16 March 1994, p. B7 Gina Kolata
Fisher's large multisite project followed 4000 women who had surgery for breast cancer. It was found that 25 of those who took tamoxifen developed uterine cancer and 6 died. Only 2 of those who did not take tamoxifen developed uterine cancer, and neither died. (We are not told how many took tamoxifen.) Previous studies had led researchers to believe that tamoxifen could cause at least a threefold increase in cancer of the uterine lining but that those cancers were not fatal and easy to treat. Experts' opinions are provided on the effect this information should have on the ongoing study of giving tamoxifen to healthy women with a high risk for breast cancer. It has caused a revision in the information given to those participating in the study. Some say that the study should be stopped. Others feel that, with properly informed subjects and the option to stop later if problems arise, the study should go on. DISCUSSION QUESTIONS: (a) Without knowing how many women in the study took tamoxifen, is it possible to determine the risk of developing uterine cancer from the information given. (b) What information would a person need to decide if it is a good idea to participate in the cancer prevention trial? (c) Do you think this trial should be stopped? <<<========<<
>>>>>==========>> Why red wine is apparently good for you. The San Francisco Examiner, 15 March 1994, p A9 Daniel Q. Haney
This article reports a possible scientific explanation for the "French paradox": the French, who eat more fat than Americans, have 75 deaths per 100,000 per year attributable to heart attack from clogging of the arteries compared with 200 deaths per 100,000 in the U.S. A University of Wisconsin study found that drinking red wine (but not white) reduces the tendency of blood to clot. Fatty meals are known to increase blood clotting, a factor in heart attacks. This was not a clinical trials study. Rather, the senior author of the study experimented on himself and six colleagues. He compared changes in the blood after drinking 2 1/2 glasses of a French red wine or a white wine from the same region. Forty five minutes after drinking the red, volunteers' blood platelets clumped together 39% more slowly than before drinking. Drinking white wine had a negligible effect. It is hypothesized that a compound called quercatin (a variety of flavonoid), contained in the skins of grapes, acts like aspirin in preventing clotting. Since white wine is made without the skins, this would explain the absence of the effect there. <<<========<<
>>>>>==========>> Anger on top of disease doubles heart attack risk. The San Francisco Examiner 19 March 1994, p. A7 Paul Raeburn,
A study has found that people with heart disease more than double their risk of heart attack when they get angry. The study was based on reports from 1122 men and 501 women who had survived heart attacks. In both groups it was found that their risks were about 2.3 times higher during the two hours after an episode in which they got angry. These conclusions are based on interviews with the patients a few days after their attacks. Researchers compared patients' behavior two hours before the attacks to their behavior the day before. Thirty-six reported being severely angry in the two hours before their attacks; only nine said they were angry the previous day. The study is considered to be of particular interest because, while many studies have identified risk factors for heart attack (e.g., high blood cholesterol, high blood pressure, smoking) there has been little research on what actually triggers an attack in an individual who is at risk. Related research is expected on whether alcohol, coffee or heavy meals can also act as triggers. <<<========<<
>>>>>==========>> The science of government gambling: another rip in the fabric of our time. Ottawa Citizen, 14 March 1994, p. A6 Jim Coyle
The U.S. News and World Report this week has a cover story title "America's Gambling Fever: How Casinos Empty Your Wallet." Coyle discusses some of the highlights of this article. Manipulation of time is said to be the most import factor in the house's attempt to maximize their winning. Special promotions are given to get the customers started early, and tricks to keep them going are employed -- an extra five minutes for each customer can mean millions to the casino. Obviously, this means that casinos spend a lot of time on proper lighting, clean air, cheap buffets etc. Dealers are under orders to issue the lowest denomination chips to players. Apparently, we might be tempted to walk away with a $25 chip but happily blow our $5 chips. In additions to a fascinating discussion of how casinos try to maximize the take, the U.S. News article has great quotes from owners and others involved in the casinos. <<<========<<
>>>>>==========>> Letter to the Editor. Chance Magazine, Winter 1994, p. 3 Peter Lacherbruch
In the Winter 1993 issue of Chance Magazine, Belin and Annsolabehere wrote an article on the margin of error in a poll. They pointed out that the newspaper determines the "margin of error" by using the 95% confidence interval for one of the percentages, ending up with about a 3% error for a sample of 1000. However, when they talk about the election, they are clearly interested in the difference between the two candidates. In the case of two candidates and no undecided voters, using the variance for the difference gives a margin of error twice as big, or 6%. When there are more than two candidates or two candidates and some undecided voters, the variance of the difference involves the covariance. The calculation of the variance leads the authors to suggest an approximation which amounts to multiplying the margin of error provided for one candidate by sqr(3), to get the margin of error for the difference. Thus a 3% error leads to about a 5% error for the difference. A letter from Peter Lachenbruch proposes another approximation for the margin of error for the difference when there are undecided voters. It is obtained by assuming that the two unknown probabilities are equal, as would be approximately correct in a close race. He shows that this is a better estimate than that obtained by sqr(3) method. The authors reply that, while they agree Lachenbruch's method gives a better estimate, they still like their own because, knowing only the margin of error, they can make the correction while watching the news. In the case where only the margin of error is given, Lachenbruch would have to solve a quadratic equation and would presumably be distracted from the TV. <<<========<<
>>>>>==========>> A James Bond with $100 tries out a tribal casino. The New York Times, 18 March 1994, p. C1 Michael T. Kaufman
Kaufman was given an bankroll of $100 to wander through the Foxwoods Resort and Casino in Leyard Connecticut to see what the opportunities are. He describes his experience in a lively manner and besides learning what he saw and how he did, you can learn how some of the less well known games are played, such as acey-deucey, chuck-a-luck and Pai Gow poker are played. <<<========<<
>>>>>==========>> Letters to the Editor. Chance Magazine, Winter 1994, p. 4 Dorothy Rice and Stanton Glantz
The Fall 1993 issue of Chance had two articles on the relationship between environmental tobacco smoke exposure and lung cancer in nonsmokers. The first article by J. Gross argued that the data does not support the claim that passive smoking causes lung cancer. The second argued that it does. Gross acknowledged support by a grant from the Tobacco Institute. The writers feel that this support leads Gross to present a biased article and, even worse, the Tobacco industry will be able to point to an article in an official journal of the American Statistical Association as evidence for their arguments. The editors reply that they recognized the issues involved but felt that the studies of passive smoke are still in the stage where there is room for debate and that it is healthy to have such debate. They encourage others to write about their opinions on this and other related editorial policies. <<<========<<
>>>>>==========>> Resizing triathlons for fairness. Chance, Winter 1994, pp. 20-25 Howard Wainer and Richard D. De Veaux
A triathlon typically has a swimming segment, a running segment and a cycling segment. How should the length of each segment be chosen if the winner is the contestant with the smallest total time. In the current big triathlons, the running distance is about 10 times the swimming distance and the cycling distance is about 40 times the swimming distance. A triathlete suggested that this choice is made to equally spread out the competitors. The amount of discrimination among competitors that can be achieved in a relatively short swimming race required a much longer cycling race. The authors admits that this might be appropriate with the present contestants but points out that the present pool of candidates may not include many good swimmers because the swimming counts so little for the total. This leads them to a discussion of self-selection bias. It is suggested that, instead of worrying about variation, the distances be chosen so that the world's fastest runner, the worlds fastest cyclist and the world's fastest swimmer would complete the segment in their specialty in about the same time. DISCUSSION QUESTIONS (a) Do you think the authors proposal would be fairer than the present system? (b) Do you have a better proposal? <<<========<<
>>>>>==========>> "Techno-thriller" statistics: chance in the fiction of Michael Crichton. Chance Magazine, Winter 1994, pp. 32-34 Allan J. Rossman
The author points out that Crichton routinely uses serious probability and statistical arguments in his novels . For example, in Jurassic Park the mathematician Malcom looks at the distribution of the heights of the Procompsognathids (scavenging little bird lizards which star in the first few chapters because of their taste for small children). He remarks that the heights have a normal distribution in contrast to what should happen since they were introduced into the park at three different times and assumed not to be able to breed. The author gives examples from six different novels and, in each case, observes that Crichton makes an interesting use of probability and statistics but, at the same time, makes a number of obvious mistakes in doing so. In the Jurassic Park example, the empirical curve is completely symmetric, very unlikely for an empirical histogram. Also, the labels on the y-axes are frequencies but, in some cases, are not integers. The author concludes that, despite his quibbles, Crichton "commendably incorporates probabilistic and statistical thinking into his writings." DISCUSSION QUESTION: In a novel, is it important to get the details of statistical or probabilistic reasoning correct. <<<========<<
>>>>>==========>> Expert opinion on climatic change. American Scientist, Jan-Feb 1994, pp. 45 William D. Nordhaus
The author wanted to see if there is any consensus among experts on global warming and its possible future effects. Through a screening process he ended up with 19 experts for his survey -- 10 economists, 4 other social scientists and 5 natural scientists. A survey was designed, asking the respondents to estimate the economic changes to be expected with major changes in the world's climate, under three different scenarios: Scenario A: A doubling of the atmospheric concentration of carbon dioxide by the mid-21st century, resulting in a 3-degree-Celsius rise by 2090. Scenario B: A continuation of that trend, reaching a six-degree warming by 2175. Scenario C: A more rapid warming in which global average temperature increases by 6 degrees by 2090. In the first set of questions, the respondents were asked to estimate the economic impact of each scenario as a percent of the gross world product estimated in 1991 to be $21.6 trillion dollars. In the second set of questions respondents were asked to indicate the uncertainties of their answers to the first questions. In addition to giving their best guesses (50th percentile) of outcomes for each scenario, they were asked to estimate the 10th percentile(smaller impact) and 90th percentile (higher impact). As a second measure of uncertainty, they were asked to give the probability of a "high consequence outcome" (defined as a lowering of global incomes by 25 percent). They were asked how this impact would be spread, depending upon the kind of goods involved or on the economic status of the region considered. Not surprisingly the author found large variability both overall and between disciplines. The author discusses reasons for this variability. The experts also included in their responses to the survey some general remarks. In the general remarks of one expert you find a question: "What will win this race between our tendency to quarrel and pollute and our power to reason and compute?" followed by: "The answer must await the roll of the dice called history." ACTIVITY: Ask the students to estimate about 25 numbers such as the distance to Washington DC, the number of students in your college, etc. Then ask them to give 90 percent confidence intervals for their estimates. See how many of their answers are within their 90 percent confidence intervals. <<<========<<
>>>>>==========>> Radon: The 'Silent Killer'. Garbage, Spring 1994, pp. 18 Leonard A. Cole
Cole argues that the EPA's aggressive campaign to persuade homeowners to test for radon gas (which, with testing and remediations costs, could cost as much as $45 billion) is not justified by scientific evidence. He states that assumptions of risk are based on a linear extrapolation from studies of miners exposed to thousands of pCi/l of radon for prolonged periods, that showed increased risk of lung cancer. This led the EPA to establish a safe level to be a level below 4pCi/l. Cole claims there is no assurance that linear extrapolation is appropriate or that one should extrapolate all the way to 0 rather than have a cut-off value. He says that the few epidemiolgy studies which have looked for a clear correlation between radon and lung cancer have given ambiguous results. Cole opposes bills, now working their way through the House and Senate, that would require every school and federal building to be tested for radon and that every home sale involving federal financial backing include a radon test. He suggests that a more reasonable approach would be to use the 20 rather than the present 4 as the "unsafe" level as, Canada does, and then try to identify the much smaller set of unsafe homes. DISCUSSION QUESTIONS: (a) How would you design a study to determine the effect of radon in homes and schools? (b) Do you agree with the philosophy "it is better to be safe than sorry" applied to a situation like this. <<<========<<
>>>>>==========>> Radon Risks are Real. Garbage, Spring 1994, p. 29 Stephen Page
Page replies to Cole for the EPA, claiming that scientific consensus is on the side of assuming a linear extrapolation and that a careful examination of the epidemiological studies suggests a real danger. Basically, while agreeing that the whole story is not yet in. Page argues it is better to be safe than sorry and mentions other areas where public policy takes this position. <<<========<<
>>>>>==========>> AIDS cases increase among heterosexuals. The New York Times, 11 March 1994 , p. A12 Lawrence K. Altman
The Centers for Disease Control and Prevention reported the number of new cases of AIDS was up 111%, mostly as a result of the new definition of AIDS. The C.D.C. had estimated that the new definition would result in a 75% increase but they said "We are in the ball park, and the difference was not surprising." (Incidentally, the headline of one of the other newspaper's account of this report was "Number of AIDS case this year doubled" The largest proportionate increase in this country is from heterosexual transmission which still accounts for a relatively small 9% of the total 103,500 cases reported in 1993. The article points out that it is difficult to accurately categorize heterosexual transmission, because it is difficult to say by which route H.I.V. was transmitted in a case, say, of an intravenous drug user who has heterosexual contacts. Apparently, the C.D.C. would count such as person by the highest risk factor, which in this case would be intravenous drug user. In the period from 1989 to 1993, the proportion of heterosexual transmissions increased from 1.6% to 9%, while the proportion of cases attributed to sex among gay men decreased from 66.5% to 46.6%. Over this same period of time, the proportion of cases attributed to injecting drugs increased, from 17.4% to 27.7% among women and heterosexual men. <<<========<<
>>>>>==========>> Surprises across the cultural divide. Science, 11 March 1994, p. 1468 Marcia Barinaga
Jim Megaw sent a questionnaire to more than 1000 physics departments around the world, asking what proportion of their faculty and students were women. It is stated that his data collected from the 400 departments who replied is one of the few studies comparing representation of women in specific scientific disciplines around the world. From Megaw's modest study, it is clear that the most industrialized countries have the smallest percentage of women physics faculty. The countries with less than 5% were Japan, Canada, W. Germany, Switzerland, Norway, Korea and the U.S. Hungary had the highest proportion with about 47%. Portugal was next with 34%, followed by the Philippines with 32% and USSR with 30%. This articles discusses some of the reasons for the differences between countries, and other articles on the same subject in this issue of Science discuss these questions. They all stress that very little has been done to really try to understand these cultural differences and think that more serious studies should be carried out. <<<========<<
>>>>>==========>> !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! CHANCE News 3.04 (6 March 1994 to 20 March 1994) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Please send suggestions to: jlsnell@dartmouth.edu >>>==========>>|<<==========<<< >>>==========>>|<<==========<<<