Confidence intervals are derived from the normal approximation to the sam-
pling distribution of lnðORÞwith variance
Var½lnðORÞF
1
a
þ
1
b
þ
1
c
þ
1
d
(ln islogarithm to base e, or thenatural logarithm.) Consequently, an approxi-
mate 95% confidence interval, on the log scale, for odds ratio is given by
ln
ad
bc
G 1 : 96
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
a
þ
1
b
þ
1
c
þ
1
d
r
A 95% confidence interval for the odds ratio under investigation is obtained by
exponentiating(the reverse log operation or antilog) the two endpoints:
ln
ad
bc
1 : 96
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
a
þ
1
b
þ
1
c
þ
1
d
r
and
ln
ad
bc
þ 1 : 96
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1
a
þ
1
b
þ
1
c
þ
1
d
r
Example 4.11 The role of smoking in pancreatitis has been recognized for
many years; the data shown in Table 4.8 are from a case–control study carried
out in eastern Massachusetts and Rhode Island in 1975–1979 (see Example
1.14). We have
(a) For ex-smokers, compared to those who never smoked,
OR¼
ð 13 Þð 56 Þ
ð 80 Þð 2 Þ
¼ 4 : 55
and a 95% confidence interval for the population odds ratio on the log
scale is from
TABLE 4.8
Use of Cigarettes Cases Controls
Current smokers 38 81
Ex-smokers 13 80
Never 2 56
166 ESTIMATION OF PARAMETERS