Studies Have Shown... Trivia Quiz

Answer: None of these.

I contacted three high-performing FunTrivia quiz takers and asked them two questions: Had they ever eaten walnuts? Did they eat walnuts for first time before they played their last quiz? Since I queried the three separately, I chose to count each contact as a separate survey in which 100% of respondents did, in fact, report consuming walnuts prior to taking quizzes. Although I might merely be accused of the statistical fallacy of "hasty generalization", use of deceptive language is, perhaps, a more accurate characterization of the claims I made in this question.

Have you started eating walnuts yet? Before you conclude that reasonable intelligence protects against rash action based on hasty, ill-founded statistical interpretations, consider the behavior of a technology company that the BBC reported. This firm clamped down on sick leave after finding that a whopping 40% of sick days we taken on Friday and Monday. Then they realized that 40% is the percent to be expected if 100% of the sick days were spread evenly over five work days (100% /5 = 20% X 2 days = 40%).
From "How To Understand Statistics" http://www.bbc.co.uk/dna/h2g2/A1091350

Answer: The number of non fatal injuries increased after tin helmets were introduced.

During World War I, statistics did show that the number of non fatal injuries increased after the replacement of cloth helmets with tin helmets. However, that's because the number of fatalities diminished, and those that had formerly been dying of their injuries swelled the ranks of the injured.

Answer: Unhelmeted motorcyclists were found to have a significantly higher incidence per collision of head injury and fatality.

Following the repeal of Louisiana's helmet law, both helmeted and unhelmeted fatalities increased. Therefore, the increase in fatalities for unhelmeted riders is not, by itself, an indication of increased risk of injury from not wearing a helmet. An increase in the fatality rate of all patients even coupled with the decrease in helmet use does not insure that the increase in fatalities occurred in the unhelmeted group. Therefore the increased fatality rate in all riders could not be attributed to not wearing helmets. That the number of head injuries and death per accident was higher in the unhelmeted group is strong evidence that not wearing a helmet is a risk factor for those outcomes. Is it a cause? Consider that in a New Mexico study ("Ann Emerg Med." 1992 Mar;21(3):279-83) unhelmeted riders involved in fatal motorcycle accidents were intoxicated 51% of the time compared to an 18% intoxication rate for helmeted riders. One could speculate that behavior induced by intoxication accounts for the increased fatality rate in accidents involving coincidentally unhelmeted riders.

The issue of legislation becomes rapidly even more complex. Even those who concede the protective benefit of helmets argue that legislation should not abridge an individual's right to assume risk. The AMA counters concerns about the cost of care that is passed on to society by uninsured riders with statistics showing the cost to be minimal in comparison to total costs of health care and/or costs attributable to accidents involving all varieties of motor vehicles. (http://www.amadirectlink.com/legisltn/positions/helmet.asp) Please note that the "AMA" here referred to is not the American Medical Association, but the American Motorcyclist Association. If it were the medical association rather than the motorcycle rider's association, would that affect the weight you gave to the argument attributed to the AMA?

Answer: My study was retrospective, and retrospective studies never provide statistically valid, useful results.

Retrospective studies utilize data gathered in the past. Retrospective studies can never demonstrate cause and effect, however statistical methods can be usefully and validly applied to data collected in a retrospective study. Well-designed and well-executed retrospective case control studies are very useful, particularly in human research. Because they rely on the results of events in the past, they do not require that study participants be subjected to risks. If a case control study reveals detrimental events in the subjects who receive an intervention, the intervention may be appropriately abandoned without participants in a prospective study experiencing harm to "prove" that the intervention is counterproductive. In other situations, anecdotal evidence may be so strong (for instance the efficacy of penicillin in pneumonia caused by susceptible organisms) and the consequences of non-treatment so severe that a prospective study would be ethically unacceptable.

Had I wished to, I could have used the following method to obtain my control group of three players. I could have identified two players with low scores on two of my own quizzes (chronic under performers). I could have read the "About" blurb on their personal FunTrivia page looking for any references to walnuts. Finding none I could report my inability to find any evidence of walnut use in these players. Oh, and the third player. That would be me. I have plenty of low scores and I don't mention walnuts in my blurb once.

Answer: Blinding

In a randomized, controlled, double blinded study, patients receive placebo or the proposed intervention without knowing (blinded) to which they are receiving. The investigators who determine the response to treatment are also blinded with respect to the actual treatment being received.

In the aforementioned study, the unblinded investigators found that patients receiving treatment benefited relative to placebo patients even though blinded investigators determined there was no difference. Patients and investigators alike are predisposed to "see" improvement unless they are "blinded". Neurology. 2001;57:S31-5.

Answer: 1

If I were to inform you that I was about to give you 50% of the money in my bank account, chances are your first question would be, "How much is in the account." The absolute risk reduction for an individual is determined by multiplying the risk of an event for that individual by the percent risk reduction for the intervention.

The risk for a 65-year-old non-smoking man with the same cholesterol and blood pressure as the woman in this question would be 14% over 10 years. The absolute risk reduction with a 50% effective treatment would then be 7%.

Answer: < 5%

Application of Bayes' theorem to the above example shows that for each positive treadmill in a woman with a 1% risk that actually had coronary artery disease, we would expect twenty two positive treadmills in similar women free of coronary artery disease. Such "free" treadmills could end up exacting quite a cost in additional testing in the 22 out of 23 women screened who were, in fact, free of coronary artery disease as well as potentially affect the women's insurability.

The pre-test probability of a condition affects the effectiveness and usefulness of a test with a particular degree of accuracy. Screening tests being performed to detect the presence of disease in populations with a low incidence of disease must be highly accurate, or most of the positive results will be false positives. In a group of patients highly likely to have coronary artery disease, even the patients having negative treadmills would still be likely to have coronary artery disease.

Answer: The difference in test scores is highly unlikely to be explained on the basis of chance variation of statistical samples.

High statistical significance refers to the fact that differences are unlikely to be attributable to chance. The actual differences may be quite small, as is the case with intelligence scores of first-born versus later born siblings.

Answer: Elderly male beagles

Study results can only be confidently applied to subjects matching the characteristics of the study group. Not only do animal studies have limited applicability to humans, results in one group of humans may not apply to differently constituted groups. Results of therapeutic interventions may often differ according to age, race or gender. For instance, certain classes of blood pressure reducing agents are more effective in black patients.

Answer: No

Under the "Dietary Supplements Health and Education Act" of 1994, my supplement will require no FDA testing. I need not even notify them of my marketing of this new product because it contains no "new dietary ingredients" that are "vitamin[s], mineral[s], herb[s], or other botanical, amino acid, or dietary substance[s] for use by man to supplement the diet." What's more, it is the Federal Trade Commission that would be responsible for regulating any advertising claims I might make for my product.

I am a perfect sucker for a deceptively done or creatively reported study. Many years ago, in the same issue of the New England Journal of medicine, a study on cholesterol lowering and a study on free health care were published. It was concluded that the small reduction in death and morbidity seen in the free care group could not justify the increased costs generated by people's "needless" visits to doctors when it cost them nothing. On the other hand, the dramatic reduction in mortality with lowering cholesterol fully justified widespread use of medication for that purpose. I fell for it, hook, line and sinker. A week later an editorial in the NEJM analyzed the raw data. Based on the published data, it would cost at least 9 times as much to save one life with cholesterol lowering medication as it would cost to save one life by providing free health care. The author pointed out that people often favored solutions that involved technology and medical intervention and often opposed giving anyone anything for nothing. I was mortified.