(a)Calculate separately for the three weight groups the odds ratio asso-
ciated with estrogen replacement.
(b)Compare the three odds ratios in part (a). When the di¤erence is
confirmed properly, we have an e¤ect modification.
(c)Assuming that the odds ratios for the three weight groups are equal
(in other words, weight is not an e¤ect modifier), calculate the
Mantel–Haenszel estimate of this common odds ratio.1.24 The role of menstrual and reproductive factors in the epidemiology of
breast cancer has been reassessed using pooled data from three large
case–control studies of breast cancer from several Italian regions (Negri
et al., 1988). In Table E1.24 data are summarized for age at menopause
and age at first live birth.
TABLE E1.24
Cases Controls
Age at first live birth
< 22 621 898
22–24 795 909
25–27 791 769
b 28 1043 775
Age at menopause
< 45 459 543
45–49 749 803
b 50 1378 1167(a)For each of the two factors (age at first live birth and age at meno-
pause), choose the lowest level as the baseline and calculate the odds
ratio associated with each other level.
(b)For each of the two factors (age at first live birth and age at meno-
pause), calculate the generalized odds and give your interpretation.
How does this result compare with those in part (a)?1.25 Risk factors of gallstone disease were investigated in male self-defense
o‰cials who received, between October 1986 and December 1990, a
retirement health examination at the Self-Defense Forces Fukuoka
Hospital, Fukuoka, Japan. Some of the data are shown in Table
E1.25.
(a)For each of the three factors (smoking, alcohol, and body mass
index), rearrange the data into a 32 table; the other column is for
those without gallstones.
(b)For each of the three 32 tables in part (a), choose the lowest level
as the baseline and calculate the odds ratio associated with each
other level.
44 DESCRIPTIVE METHODS FOR CATEGORICAL DATA