Appleton, French, and Vanderpump (1996) present data that would appear to show that smoking is good for you. They assessed smoking behavior in the early 1970s and then looked at survival data 20 years later. For simplicity they restricted their data to women who were current smokers or had never smoked. Out of 582 smokers, 139 (24%) had died. Out of 732 smokers, 230 (31%) had died. Obviously proportionally fewer smokers died than nonsmokers. But if we break the data down by age groups we have Age Smoker 18–24 S NS 25–34 S NS 35–44 S NS 45–54 S NS 55–64 S NS 65–74 S NS 75+

S NS Dead 2 1 3 5 14 7 27 12 51 40 29 101 13 64 Alive 53 61 121 152 95 114 103 66 64 81 7 28 0 0

a. Explain these results as an example of Simpson’s paradox.

b. Apply the Mantel-Haenszel test to these data.