Freshman Statistics Seminar
Week 7:Example leading to Statistical tests
Objective: To look at how evidence can be refuted and why. Also, to look at more sophisticated evidence to overcome objections.
- Pearl 1938 Science 87 p216 “Tobacco Smoking And Longevity”
The first article is a look at the mortality rates for smokers versus nonsmokers for a registry of records from the laboratory.
- Doll 1954 Br Med J 2 p1451 “The Mortality Of Doctors In Relation To Their Smoking Habits- A Preliminary Report”
The second article is from 16 years later, and it looks at doctors across England who differ primarily on smoking habits.
- Doll 2004 BMJ 328 p1519 “Mortality In Relation To Smoking: 50 Years Observations On Male British Doctors”
The third article is a 50-year follow up to the second article, which adds the additional element of following over time.
Suggested Lesson Structure:
Perform the active learning module, with perhaps a couple of follow-up questions.
Active Learning Modules:
Divide the class up into small groups (5-7 students per group). Each group is a group of tobacco company scientists who are charged with finding multiple reasons why there seems to be an increase in deaths in the tobacco use group BESIDES tobacco use. The groups need to come up with several each. Each group will need to name one different potential confounder, and the groups’ contributions will be written on the board. This will (hopefully) lead to a lot of creativity, and also, willingness to be an early contributor.
Now, the groups are should try to refute the confounders they gave in class by using the two BMJ articles. Alternatively, this can be a writing assignment to be turned in the next class.
- Does science drive statistical advances, or the other way around?
- Why can’t we do a randomized controlled trial?
Smoking and Health
Study Group on Smoking and Health
Science, New Series, Vol. 125, No. 3258 (Jun. 7, 1957), pp. 1129-1133