Confidence Intervals for

Proportions
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
In a 1999-2000 longitudinal study of youth baseball, researchers found that 77 of 248 young
pitchers complained of elbow pain after pitching. What is the 90% C.I. for the true population
proportion of young pitchers with elbow pain?
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

𝑝^ ± 𝑧 𝛼/ 2
√ 𝑝^ (1− 𝑝^ )
𝑛

Conditions for inference:

• and
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

𝑝^ ± 𝑧 𝛼/ 2
√ 𝑝^ (1− 𝑝^ )
𝑛

We know and . We need to find , which depends on the

desired confidence level.
 3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

What value of z marks the lower tail area of 0.05?

0.90

data cl90;
1 − 0.90 z90 = quantile('normal',0.05);
=0.05
2 run;

proc print data=cl90;

run;
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

√
𝑝^ ± 𝑧 𝛼/ 2
𝑝^ (1− 𝑝^ )
𝑛

0.3105 ± 1.645
√
0.3105(1 −0.3105)
248

0.3105 ± 1.645
√
0.3105(0.6895)
248
0.3105 ± 1.645 ∗ 0.029381

Margin of
0.3105 ± 0.048332
Error
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

^ =31.05 % ; 90 %  𝐶𝐼 (26.22 % , 35.88 %)

𝑝

Interpretations
• We are 90% confident that the true population
proportion of elbow pain among young pitchers is
between 26.22% and 35.88%.
3.1 C.I. AROUND ONE POPULATION
PROPORTION (p) [using sample statistics]:
What is the 90% C.I. for the true population proportion of young pitchers with elbow pain?

data zci;
x=77;
n=248;
p=x/n;
𝑝^ ± 𝑧 𝛼/ 2
√ 𝑝^ (1− 𝑝^ )
𝑛

alpha=1-0.9;
z=quantile('normal',1-alpha/2);
se=sqrt((p*(1-p))/n);
L=p-z*se;
U=p+z*se;
run;

proc print data=zci;

var p L U;
format _numeric_ percent9.2;
run;
Confidence Intervals
• Confidence intervals that use the normal approximation have the
following form: