!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
CHANCE News 2.17
(3 Oct. to 20 Oct. 1993)
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
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 and full
text of the newspaper articles can be found on the
chance gopher. Just point your gopher to
chance.dartmouth.edu
===================================================
If a man will begin with certainties he shall end up in
doubts; but if he will be content to begin with doubts
he shall end in certainties.
Francis Bacon
===================================================
ARTICLES ABSTRACTED
>>>>>==========>> Gudmund Ivererson and Norton Starr suggested the following: Using Steinbeck, Hemingway and Others as Parole Officers. The New York Times, 6 Oct 1993, B10 Special to The New York Times This article reports on a Massachusetts program to influence convicts - a class for convicts held on the Dartmouth campus of the University of Massachusetts. The alternative sentencing program is for those who have been convicted of crimes ranging from burglary to assault; repeat offenders are eligible. To be accepted, convicts must be literate and able to convince a judge and probation officer that they want to turn their lives around. The program puts to the test the concept that great literature can influence lives. A recent study of the program's results was encouraging. C. Roger Jarjoura, a professor of criminal justice at Indiana University, found that people who took the class were less likely to be convicted of new crimes than people who had not. Professor Jarjoura said he followed 32 men who took the class for one to two and a half years and found that six of them were convicted of new crimes. Over the same periods he tracked 40 men with similar backgrounds and records who did not take the class; 18 of them were convicted of new crimes. Not, apparently, a fair comparative experiment, but intriguing nevertheless. <<<========<<
>>>>>==========>> David Lloyd suggested this next item: The role of p-values in analyzing trial results. Statistics in Medicine, Vol 12, 1443-1452, 1993 Peter R. Freeman The author explains why he feels that the use of p- values as the principal method assessing and reporting results of clinical trials cannot be defended. He starts with the results of a multiple choice question on a survey given to doctors, dentists and medical students. The survey asked them which of four descriptions they preferred for the meaning of: A controlled trial of a new treatment led to the conclusion that it is significantly better than placebo (p < .05). Which of the following statements do you prefer? 1. It has been proved that treatment is better than placebo. 2. If the treatment is not effective, there is less than a 5 percent chance of obtaining such results. 3. The observed effect of the treatment is so large that there is less than a 5 percent chance that the treatment is no better than placebo. 4. I do not really know what a p-value is and do not want to guess. The proportions of 397 responders for the four options were 15%, 19%, 52% and 15% respectively. Since there was only about a 50% response rate the assumption is that this is biased in favor of the informed group. Freeman explains why option 3 may not be as bad as it seems but then goes on to tell us why he still is pretty discouraged about the use of p-values and what he thinks can be done to change how results of clinical trials are reported. The same issue of this journal has other articles and discussions on the Bayesian versus the frequentist approach to the interpretation of clinical trials. <<<========<<
>>>>>==========>> Bob Abelson suggested that the following article has a lot of possibilities for discussion which indeed it does. High murder rate for women on job. The New York Times, 3 Oct 1993, Sec 1 page 29 AP This article starts off by stating that the Government's first comprehensive accounting for how people die at work showed that 40 percent of the women who die on the job are murder victims while only 15 percent of the men who die at work are murdered. Officials suggested that the difference was due to the large number of women exposed to crime while working late at night. As astute readers noticed, the additional statistics provided in this article show that the headline and these remarks are misleading. One reads that men accounted for 93 percent of all job- related deaths, even though they constitute only 55 percent of the work force. Thus of the 6,083 workers who the Labor Department reported as having died on the job in 1992, 426 were women and 5,657 were men. The point is that men dominate in the most dangerous occupations -- construction workers, farming-forestry workers, transportation workers, and protective workers such as police and firemen. 40 percent of the 426 or 170 women who died on the job were murdered. 15 percent of the 5,657 or 849 men who died on the job were murdered. Looked at this way, about five times as many men were murdered as women even though there were almost as many women as men on the work force. What would be a good way to compare the rate of women murdered with the rate of men murdered in the work force? Those who commented on this article generally blamed the Labor Department for too zealous an attempt to demonstrate that women are victimized. <<<========<<
>>>>>==========>> On not misreading murder statistics. The New York Times, 10 Oct 1993, Sec 4, Page 14 Letter to the Editor from Kingsley R. Browne A letter calling to the attention readers of the New York Times of the misleading comparison of murder rates on the job for men and women made in the previous article. <<<========<<
>>>>>==========>> Mozart makes the brain hum, a study finds. The New York Times, 14 Oct 1993, B9 Malcolm W. Browne A report on a letter in the current issue of Nature that describes an experiment to see if listening to music before taking an IQ test improved a students performance on the test. 36 students each took three different spatial IQ tests under three different conditions: (a) listening to 10 minutes of Mozart's Sonata for Two Pianos in D Major (b) listening to relaxing exercises for ten minutes, (c) ten minutes of silence. The scores for those who listened to Mozart were signficantly higher than under the other two options. The results under the other options were quite similar. The authors plan to do similar tests for different kinds of music. They conjecture that some kinds of music may, in fact, interfere with, rather than enhance, abstract reasoning. We had good luck with this as a discussion topic for our CHANCE course. <<<========<<
>>>>>==========>> Oh baby! January's storm produced a flood of births. Seattle Times, 19 October 1993, A1 Mary F. Pols A power shortage occurred on Puget Sound on Jan. 20, 1993 that lasted almost five days. Now about forty weeks later this article looks at number of births in the local hospitals. At Valley Medical Center it is said that the average for births is about 7 a day, but 11 births were recorded on both Friday and Saturday of last week. At Bellevue's Overlake Hospital 127 babies were born between Sept. 30 and Oct. 14, up from 96 last year and 97 in 1991. Of course, this is reminiscent of the famous New York blackout where there were thought to be too many Monday birthdays at the appropriate time after the blackout. It has been suggested that this was, in fact just the well know cyclic behavior for birthdays with a maximum on Mondays. Perhaps legends produce legends. <<<========<<
>>>>>==========>> Sunscreen shown to help prevent cancer. Los Angeles Times, 14 Oct 1993, A21 Times staff and wire reports Australian researchers report in the current issue of the New England Journal of Medicine on a study showing for the first time that using sunscreen helps prevent skin cancer for those at the highest risk. The study was a randomized study with 588 men and women all of whom had solar keratoses that are the forerunners of skin cancer. Those who used SPF-17 averaged a net loss of about a half a keratosis, while those on placebo gained one. About 700,000 Americans each year are diagnosed with easily treated skin cancers. About 32,000 get melanoma that can spread quickly and is deadly unless recognized early. The incidence of melanoma has increased consistently over the past 20 years and felt to be because of increased sun exposure. <<<========<<
>>>>>==========>> Gun in home? Study finds it a deadly mix. The New York Times, 7 Oct 1993, A18 Warren E. Leary The New England Journal of medicine this week reports on a study that shows that a handgun in a home almost triples the chances that someone will be killed there. The researchers looked at homicides from August 1987 to August 1992 in three metropolitan areas. Of 420 homicides that occurred in the home of the victim about 77% were by a spouse, family member or someone the victim knew. About half of the deaths were due to gunshot wounds. Even in the 14 percent of the cases involving forced entry a gun in the home was found to be of little protection. The researches found that homicide in the home was more likely if there was a history of alcohol or illicit drugs or of domestic violence. Even controlling for factors like these the risk of homicide was about 2.7 times greater in homes that had guns. The National Rifle Association criticized the findings. <<<========<<
>>>>>==========>> Cancer study error found. Boston Globe, 6 Oct 1993 Alison Bass A study of about 120,000 women in the Nurses Study published in July 21 issue of the Journal of the American Medical Association found that only 2.5 percent of breast cancer occurred in women with a family history of the disease. An error was made in this calculation caused by an incorrect transcription from tables. When corrected, the correct estimate was 6 percent which was more in agreement with previous studies. Various experts are interviewed and comment that peer review is not infallible, the media gives too much attention to a single study, etc. A recent study in Utah found that 17 to 19 percent of breast cancer in the population occurred in women with a family history of the disease. The Utah researchers had access to records going back seven generations and were able to include data on distant relations. The Nurses Study was much more limited in the ability to go back in the family and it is suggested that the best estimate is probability somewhere between the 6 and 17 percent figures. <<<========<<
>>>>>==========>> 2nd study targets ultrasounds. Newsday, 8 Oct 1993, News p 17 Anita Flanagan This is the second study whose results have led the researchers to recommend that ultrasound treatments be restricted to those who have a good reason to have them. A previous study suggested that ultrasound treatments increased the chance of having a left-handed baby. This study was designed to see if giving multiple scans would be more beneficial than giving just one and would result in less premature births etc. While it was not in the design to test the effect on weight of the babies, they did notice that those who received multiples scans had babies who weighed 1 percent less than those with a single scan. The authors are aware of the dangers of looking at things that were in the design but still suggest that the result was significant and should be looked at further. <<<========<<
>>>>>==========>> Causation and disease: A chronological journey. Alfred S. Evans Plenum, 1993,$35 ISBN 0-306442833 This book received an enthusiastic review in the Oct 9 issue of Lancet. It discusses the question of establishing causality of a disease. In particular, Evans discusses in detail the controversy over whether HIV causes AIDS, reviewing the Duesberg theories, etc. The reviewer says that it is a particularly well written-book giving an evenhanded approach to some controversial ideas. <<<========<<
>>>>>==========>> It figures: income 'jump' not real. Valley News, 22 Oct 1993 A1 Dan Billing In March of this year the Census Bureau interviewed 57,400 households in 729 sample areas for the annual survey. People were asked to fill out detailed questionnaires about their sources of income. The Census Bureau reported that the median household income remained virtually the same for the country as a whole in 1992 except for Vermont and New Hampshire that reversed prior declines and made strong increases in household income. The median household income in New Hampshire rose 10 percent to $39,644 last year and in Vermont it rose 12.6 percent to $32,829. This with adjustment for a 3-percent inflation rate. This article reports the results of surveying local industries such as Dartmouth College etc., and failing to find evidence for this large jump. The article explains how this results could come about: changes in the numbers of people in a household, a previous estimate that was too low, etc. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! CHANCE News 2.17 (3 Oct. to 20 Oct. 1993) !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Please send suggestions to: jlsnell@dartmouth.edu >>>==========>>|<<==========<<<
>>>==========>>|<<==========<<<