In my iteration of Chance this unit followed an initial unit on probability and coincidences. It could easily be the first unit in the course since it did not depend at all on the first unit. I began the unit with preliminary reading from the two statistics texts, Moore and FPPA. We actually did and handed in some homework problems, which we discussed in class. Perhaps this step could be skipped by using the journal idea of the Princeton course.
We had a short class exercise taken from the interesting little book Rival Hypotheses by Schuyler W. Huck and Howard M. Sandler (Harper and Row, 1979) from which I include the exercise and a relevant appendix on rival hypotheses. The hypothesis that the T.V. ad is trying to prove is that Orville's popcorn pops up better. I had my students work in small groups and devise rival hypotheses to this one. They did this without the benefit of the formal structure provided by the appendix, but rather based solely on their recently acquired knowledge of statistics, but the discussion was a fruitful one.
The class next discussed the articles on the aspirin studies. This is an excellent article for sharpening the students sensitivity to design issues as both studies were ``designed'', but the results differed. The students saw from this even-handed article that scientific truth most often does not readily divulge her secrets. Both studies were carefully planned but neither was without flaw. The article also raises explicitly the issue of how the press should deal with new scientific findings and this is a subject the students tend to have an opinion about. (Note: I had my students read the short book News and Numbers by Victor Cohn (Iowa State University Press, 1989). I found this book to be too redundant with the too basic statistics texts mentioned above to recommend its inclusion in future iterations of the course.)
We next discussed the smoking article to make it clear that many scientific findings necessarily come from non-experimental work. (The methodology of randomized experimentation is less that a hundred years old.) It was useful to bring to this discussion Moore's short section (section 3, chapter 5, pages 276and 277) on establishing causastion in a non-experimental setting given that a relationship between A and B has been established: (1) The association between A and B recurs in different circumstances; this reduces the chance that it is due to confounding. (2) A plausible explanation is available showing how A could cause changes in B. (3) No equally plausible third factor exists that could cause changes in A and B together. Certainly the smoking and lung cancer issue passes this test.
Finally we discussed the Channel One controversy. Channel One is a commercial News show produced by Whittle Communications to be shown in schools daily. Participating schools get free electronics equipment by agreeing to show all students these daily 12 minute programs complete with two minutes of commercial messages. The issue of Channel One has been hotly debated around the country and I rightly guessed that a number of my first year students would have been exposed to it. (I was right; 5 of the 12 students had been at schools with Channel One.) This study is quasi- experimental and made for lively class debate. I must warn you that there was a tendency for students to want to debate this issue more broadly than just whether the enclosed study was valid.