Vous êtes sur la page 1sur 6

Statistical Tests for Population Proportions

STAT 400
July 27, 2016

Example 1 A scientist wishes to test if a new treatment has a better cure rate than the traditional
treatment which cures only 60% of the patients. In order to test whether the new treatment is
more effective or not, a test group of 20 patients were given the new treatment. Assume that each
personal result is independent of the others.

We want to decide on a cure rate of p = 0.60 vs. p > 0.60.

a. If the new treatment has the same success rate as the traditional, what is the probability
that at least 14 out of 20 patients will be cured?

This is a binomial problem for P (X 14|p = 0.60),

P (X 14|p = 0.60) =

20
X
i=14

20
0.60i 0.4020i = 0.25
i

(1)

b. Suppose that 14 out of 20 patients in the test group were cured. Based on the answer for
part above, is there a reason to believe that the new treatment has a better cure rate than

If p = 0.60, then 25% of all possible samples would have 14 or more patients cured (out
of 20). Thus, it is not unusual to see 14 out of 20 patients cured for a treatment that
cures 60% of the patients. We have no reason to believe that the new treatment has a
better cure rate than the traditional treatment if X = 14.

STAT 400

Tests for Proportions

c. If the new treatment has the same success rate as the traditional, what is the probability
that at least 17 out of 20 patients will be cured?

We want to know P (X 17|p = 0.60),

P (X 17|p = 0.60) =

20
X
i=17

20
0.60i 0.4020i = 0.016
i

(2)

d. Suppose that 17 out of 20 patients in the test group were cured. Based on the answer above,
is there a reason to believe that the new treatment has a better cure rate than the traditional
treatment?

If p = 0.60, then only 1.6% of all possible samples would have 17 or more patients
cured (out of 20). Thus, it is fairly unusual to see 17 out of 20 patients cured for a
treatment that cures 60% of the patients. We have a good reason to believe that the
new treatment has a better cure rate than the traditional treatment if X = 17.

Example 2 Just prior to an important election, in a random sample of 749 voters, 397 preferred
Candidate Y over Candidate Z. Is there enough evidence to claim that over half of all the voters
prefer Candidate Y over Candidate Z?

Note that p =

397
749

Binomial can be used. The test statistic is,

p p0
Z=q

p0 (1p0 )
n

0.53 0.50
=q
= 1.64
0.50(10.50)
749

(3)

Stepanov, Culpepper

STAT 400

The critical value is Z = Z0.10 = 1.2816

d. Perform an appropriate test at a 10% level of significance ( = 0.10). What do you conclude?

We compare our test statistic to the rejection region and in this case Z > Z , so we
reject H0 because 1.64 > 1.2816.

e. Perform an appropriate test at a 1% level of significance ( = 0.01). What do you conclude?

We compare our test statistic to the rejection region and in this case Z > Z , so we fail
to reject H0 because 1.64 < 2.3263.

The p-value is P (Z > 1.64) = 1 0.9495 = 0.0505.

g. Using the p-value, state your decision (Accept H0 or Reject H0 ) at = 0.03 and at = 0.07.

Note that P (z > 1.64) = 0.0505, so the p-value > 0.03 (so we fail to reject the null)
and the p-value < 0.07 (so we reject the null).

Stepanov, Culpepper

STAT 400

Tests for Proportions

Example 3 Alex wants to test whether a coin is fair or not. Suppose he observes 477 heads in 900
tosses. Let p denote the probability of obtaining heads.

Note that p =

477
900

p p0
Z=q

p0 (1p0 )
n

0.53 0.50
=q
= 1.80

(4)

0.50(10.50)
900

c. What is the rejection region for = 0.10?

The critical value is Z/2 = Z0.05 = 1.6449, so the rejection region includes Z values
that are Z > 1.6449 and Z < 1.6449.

d. Perform an appropriate test at a 10% level of significance ( = 0.10). What do you conclude?

We compare our test statistic to the rejection region and in this case Z > Z/2 , so we
reject H0 because 1.80 > 1.645.

f. Using the p-value, state your decision (Accept H0 or Reject H0 ) at = 0.05.

Note that P (z > 1.80) = 0.0718, so the p-value > 0.05 and we fail to reject the null
hypothesis.

Stepanov, Culpepper

STAT 400

Tests for Proportions

Example 4 A certain automobile manufacturer claims that at least 80% of its cars meet the tough
new standards of the Environmental Protection Agency (EPA). The EPA tests a random sample of
400 its cars. Let p be the proportion of the cars that meet the new EPA standards. Suppose that
300 of the 400 cars in our sample meet the new EPA standards.

Note that p =

300
400

p p0
Z=q

p0 (1p0 )
n

0.75 0.80
=q
= 2.50

(5)

0.80(10.80)
400

c. What is the rejection region for = 0.05?

The critical value is Z = Z0.05 = 1.6449, so the rejection region includes Z values
that are Z < 1.6449.

d. Perform an appropriate test at a 5% level of significance ( = 0.05). What do you conclude?

We compare our test statistic to the rejection region and in this case Z = 2.50 <
Z = 1.6449, so we reject H0 .

f. Using the p-value, state your decision (Accept H0 or Reject H0 ) at = 0.001.

Note that P (z < 2.50) = 0.0062 > 0.001, so we fail to reject the null hypothesis.

Stepanov, Culpepper

STAT 400

Tests for Proportions

Example 5 In a comparative study of two new drugs, A and B, 120 patients were treated with
drug A and 150 patients with drug B, and the following results were obtained. We wish to test
whether drug B has a higher cure rate than drug A.
Drug A

Drug B

Cured

78

111

Note Cured

42

39

Total

120

150

a. What hypothesis are we interested in testing?

H0 : pA pB 0 vs. H1 : pA pB < 0

Note that pA =

78
120

= 0.65, pB =

111
150

= 0.74, and p =

78+111
120+150

is,
Z=r

pA pB
p (1 p)

=r

1
nA

1
nB

0.65 0.74
0.70 (1 0.70)

1
120

1
150

= 1.60

(6)

c. Using the p-value, state your decision (Accept H0 or Reject H0 ) at = 0.05.

Note that P (z < 1.60) = 0.0548 > 0.05, so we fail to reject the null hypothesis.

Stepanov, Culpepper