On 6 Dec 2002 13:59:13 -0800, [EMAIL PROTECTED] (Simon, Steve, PhD) wrote: [ snip] > > Perhaps we need to trumpet some of the many success stories in statistics. > One of my favorites is how careful research in the 50's and 60's established > a causal link between smoking and lung cancer.
- Is there a book or article that tells that story comprehensively? > Each individual study may > have had some flaws, but the cumulative evidence across the large number of > studies provided a level of proof that was very strong. - But "cumulative evidence" is such a bland way to put it! How far back were they called 'coffin nails'? The OED has: "1888 Texas Siftings 18 Feb. 8/1 A youth..puffed at an ill-smelling coffin nail. " In the 1950s, Fisher was still arguing that the technical requirements of proper logic had not been met, and I think he was right -- All of those studies had been samples-of- opportunity, and no one demolished *all* the alternative explanations. "Psychological factors that cause smoking could also cause cancer." I believe, though I'm not sure, that this tobacco debate is what led to the formalization of those published tenets of "epidemiological proof". That includes such things as parallel evidence from studies in-vivo and in-vitro; and mortality consistent with a dose-response. In the end, I thought that the magnitude of the effect size was one thing that was persuasive, for smoking. If the Smoking/ Lung-cancer OR was 6-fold or so, that could not be accounted for by some 'cause' that was much weaker, statistically. And: psychological factors may have a slight relationship with the quality of outcomes from diseases, but dispositions or personality have never predicted (Is there any disease with an OR of 2?) the new presence of diseases. > > I'd be interested in how other people approach this. We need a way to teach > students to have healthy skepticism, but not cynicism. -- Rich Ulrich, [EMAIL PROTECTED] http://www.pitt.edu/~wpilib/index.html . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
