Chapter 3 Hypothesis Testing: One Sample Tests

3 HYPOTHESIS
TESTING:
ONE SAMPLE
TESTS
Objectives
After studying this chapter you should

understand when, and be able, to carry out a one sample t-test;

be able to carry out a test for a normal population variance;

be able to test hypotheses for a binomial population proportion,

either by calculating exact probabilities or by using an
appropriate approximation;

be able to test hypotheses for the mean of a Poisson population

parameter, either by calculating exact probabilities or by using
a normal approximation;

be able to carry out a sign test for a population median, either

by using a binomial distribution or an approximation.

3.0

Introduction

In Chapter 10 of Statistics you were introduced to some important

basic concepts of hypothesis testing:
Null hypothesis (H0):

an assertion that a parameter in a

statistical model takes a particular value,
and is assumed true until experimental
evidence suggests otherwise.

Alternative
hypothesis (H1):

expresses the way in which the value of

a parameter may deviate from that
specified in the null hypothesis, and is
assumed true when the experimental
evidence suggests that the null
hypothesis is false.

Type 1 error:

rejecting the null hypothesis when it is,

in fact, true.

37

Chapter 3 Hypothesis Testing: One Sample Tests

Type 2 error:

accepting the null hypothesis when it is,

in fact, false.

Test statistic:

a function of a sample of observations

which provides a basis for testing the
validity of the null hypothesis.

Critical region:

the null hypothesis is rejected when a

calculated value of the test statistic lies
within this region.

Critical value:

the value which determines the

boundary of the critical region.

Significance level ( ): the size of the critical region; the

probability of a Type 1 error.
One-tailed test:

the critical region is located wholly at

one end of the sampling distribution of
the test statistic; H1 involves < or > but
not both.

Two-tailed test:

the critical region comprises areas at

both ends of the sampling distribution
of the test statistic; H1 involves .

Example
A consumer group, concerned about the mean fat content of a
certain grade of steakburger submits to an independent laboratory
a random sample of 12 steakburgers for analysis. The percentage
of fat in each of the steakburgers is as follows.
21 18 19 16 18 24 22 19 24 14 18 15

The manufacturer claims that the mean fat content of this grade
of steakburger is less than 20%. Assuming percentage fat
content to be normally distributed with a standard deviation of 3,
carry out an appropriate hypothesis test in order to advise the
consumer group as to the validity of the manufacturer's claim.
Solution
For this problem, denoting the percentage fat content by X, then

X ~ N ,32 and it is required to test

H0: = 20%
H1: < 20% (one-tailed)

0.05
5%

Accept HH00

Significance level, = 0.05 (say)

Critical region,

z < 1.645

3
X ~ N 20,
12
2

Under H0,
38

1.645
Critical region
reject HH00

Chapter 3 Hypothesis Testing: One Sample Tests

Now the test statistic is

z=

Calculation gives

x=

and thus

z=

x 228
=
= 19
12
n
19 20
= 1.15
3
12

This value does not lie in the critical region. Thus there is no
evidence, at the 5% level of significance, to support the
manufacturer's claim.

3.1

Normal population mean

(n small, unknown)

Consider again the previous example on steakburgers and now

suppose that it can be assumed that the percentage fat content is
normally distributed but with an unknown standard deviation.

Now from Section 2.1 you have seen that if X ~ N , 2 then

x

n
is a t statistic with degrees of freedom v = n 1, where
2
1
1 2 ( x )
2
2 =
(x x) =
x
.
n
n 1
n 1

So a z-test, used when is known, is replaced by a t-test when

is unknown.
What is meant by the term 'degrees of freedom' and why is its
value important?

39

Chapter 3 Hypothesis Testing: One Sample Tests

A re-working of the steakburger problem, assuming that is

now unknown, is thus as follows.
H0: = 20%
H1: < 20% (one-tailed)
Significance level, = 0.05 (say)
Degrees of freedom, v = n 1 = 11
Critical region, t < 1. 796
Under H0,

2
X ~ N 20,
12

Now the test statistic is

t=

1
2282
with x = 19 and 2 = 4448
= 10.545 and so
12
11

= 3.25.
Hence

t=

19 20
= 1.07
3.25
12

This value does not lie in the critical region. Thus there is no
evidence, at the 5% level of significance, to support the
manufacturer's claim.

Activity 1

Weights and measures

Collect a random sample of between 5 and 10 'identical' items on

which is stated the contents.
For example, boxes of matches, boxes of paper clips, packets of
drawing pins, packets of crisps, chocolate biscuits, etc.
Count or measure, as appropriate, the contents of each item.
Investigate the hypothesis that your sample mean differs
significantly from the value stated.

40

0.05
5%

Accept HH00

1.796
Critical region
H00
reject H

Chapter 3 Hypothesis Testing: One Sample Tests

Exercise 3A
1. During a particular week, 13 babies were born in
a maternity unit. Part of the standard procedure
is to measure the length of the baby. Given
below is a list of the lengths, in centimetres, of
the babies born in this particular week.
49 50 45 51 47 49 48 54 53 55 45 50 48

Assuming that this sample came from an

underlying normal population, test, at the 5%
significance level, the hypothesis that the
population mean length is 50 cm.
(AEB)
2. A random sample of 12 steel ingots was taken
from a production line. The masses, in
kilograms, of these ingots are given below.
24.8

30.8

28.1

24.8

27.4

22.1

24.7

27.3

27.5

27.8

23.9

23.2

Assuming that this sample came from an

underlying normal population, investigate the
claim that its mean exceeds 25.0 kg.
(AEB)
3. A random sample of 14 cows was selected from a
large dairy herd at Brookfield Farm. The milk
yield in one week was recorded, in kilograms,
for each cow. The results are given below.
169.6 142.0 103.3 111.6 123.4 143.5 155.1
101.7 170.7 113.2 130.9 146.1 169.3 155.5

Stating clearly any distributional assumptions

that you make, investigate the claim that the
mean weekly milk yield for the herd is greater
than 120 kg.
(AEB)

4. A random sample of 15 workers from a vacuum

flask assembly line was selected from a large
number of such workers. Ivor Stopwatch, a
work-study engineer, asked each of these workers
to assemble a one-litre vacuum flask at their
normal working speed. The times taken, in
seconds, to complete these tasks are given below.
109.2

146.2

127.9

92.0

91.1

109.8

114.9

115.3

108.5
99.0

112.8

130.7

141.7

122.6

119.9

Assuming that this sample came from an

underlying normal population, investigate the
claim that the population mean assembly time is
less than 2 minutes.
(AEB)
5. In processing grain in the brewing industry, the
percentage extract recovered is measured. A
particular brewing introduces a new source of
grain and the percentage extract on 11 separate
days is as follows.
95.2 93.1 93.5 95.9 94.0 92.0
94.4 93.2 95.5 92.3 95.4

Test the hypothesis that the true mean percentage

extract recovered is 95.0. What assumptions
have you made in carrying out your test? (AEB)
6. A car manufacturer introduces a new method of
assembling a particular component. The old
method had a mean assembly time of 42 minutes.
The manufacturer would like the assembly time
to be as short as possible, and so he expects the
new method to have a smaller mean. A random
sample of assembly times (minutes) taken after
the new method had become established was
27 39 28 41 47 42 35 32 38

Stating any necessary distributional assumptions,

investigate the manufacturer's expectation.
(AEB)

3.2

Normal population variance

In Section 2.2 you have seen that if 2 denotes the variance

[divisor ( n 1) ] of a random sample of size n from a normal
population with variance 2 , then

( n 1) 2 ~ 2
2

with degrees of freedom, v = n 1.

41

Chapter 3 Hypothesis Testing: One Sample Tests

The statistic

( n 1) 2 may thus be used to test hypotheses

2

concerning a normal population variance or, by implication, a

normal population standard deviation.

How does the 2 distribution differ in shape from the

t-distribution and how does this affect the percentage points
of the 2 distribution?

Example
A user of a certain gauge of steel wire suspects that the standard
deviation of its breaking strength, in newtons (N), is different
from the value of 0.75 as specified by the manufacturer.
Consequently the user tests the breaking strength of each of a
random sample of nine lengths of wire and obtains the following
results.
72.1 74.5 72.8 75.0 73.4 75.4 76.1 73.5 74.1

Assuming breaking strength to be normally distributed, test, at

the 10% level of significance, the manufacturer's specification.
Solution
H0: = 0. 75 N or H0: 2 = 0.5625N 2
H1: 0. 75 N or H1: 2 0.5625N 2 (two-tailed)
Significance level, = 0.10

5%
0.05

Degrees of freedom, v = 8

0 2.733

Critical region, < 2. 733 or > 15.507

2

( x )2
( n 1) 2 = x 2
n

= 49430. 49

2 =

666.9 2
= 13.2
9

( n 1) 2 = 13.2 = 23.47
2

0.5625

This value does lie in the critical region. Thus there is evidence,
at the 10% level of significance, to dispute the manufacturer's
specification as regards variability of breaking strength.
42

15.507

Calculation gives

Under H 0 ,

5%
0.05

Accept H00

Critical region
reject H00

Chapter 3 Hypothesis Testing: One Sample Tests

Activity 2

Random standardised normal

deviates

Obtain from tables, calculator, or computer a random sample of

between 10 and 15 values of the standardised normal random
variable.
Investigate the hypothesis that your sample provides significant
evidence that the variance differs from unity.

Exercise 3B
1. Given the data in Question 1 of Exercise 3A,
test, at the 5% level of significance, the
hypothesis that the population standard deviation
of length is 2.5 cm.
(AEB)
2. Referring to Question 2 of Exercise 3A,
investigate the claim that the population standard
deviation is greater than 2 kg. (AEB)
3. Using the data given in Question 3 of
Exercise 3A, test the hypothesis that the standard
deviation of yield for the herd at Brookfield
Farm is 20 kg.
(AEB)

5. With reference to Question 5 of Exercise 3A, the

variance of percentage extract recovered was 6.0
prior to introducing the new source of grain.
Does the new source of grain result in a decrease
in the variability of percentage extract
recovered?
(AEB)
6. Referring to Question 6 of Exercise 3A, the
manufacturer assumes that the new method of
assembly will reduce the variability of assembly
times. Given that the old assembly method had a
standard deviation of 8 minutes, investigate the
manufacturer's assumption.
(AEB)

4. The variability in unit mass is very critical in the

pharmaceutical industry. The mass, in grams, of
each of 15 randomly selected units of a specific
drug is as follows.
3.48 3.52 3.50 3.47 3.49 3.54 3.51 3.52
3.46 3.45 3.55 3.48 3.51 3.52 3.50

Assuming that this sample comes from an

underlying normal population, test the
hypothesis that the variance of the population
exceeds 0.002 gm 2 .

3.3

(AEB)

Binomial population
proportion

In situations where the population parameter of interest in a

hypothesis test is a proportion, or percentage, rather than a mean
or variance, then a binomial distribution, or an appropriate
approximation, provides the basis for the test.
Remember from Chapter 5 of Statistics, that if X denotes the
number of successes in n repeated trials, each of which may
result in a success with probability p,

43

Chapter 3 Hypothesis Testing: One Sample Tests

then

X ~ B( n, p )

and

n
n x
P( X = x ) = p x (1 p )
x

where

n
n!
=
x x!( n x )!

x = 0, 1, 2, K, n

Example
Until recently, an average of 60 out of every 100 patients have
survived a particularly severe infection. When a new drug was
administered to a random sample of 15 patients with the
infection, 12 survived. Does this provide evidence that the new
drug is effective?
Solution
Let X denote the number of patients who survive.
Then X ~ B(15, p) , and it is required to test
H0: p = 0.6

(not effective)

H1: p > 0.6

(effective; one-tailed)

Now assuming H0 is true, X ~ B(15, 0.6 ) , and so

15
P( X = 12 ) = 0.612 0. 4 3 = 0.06339
12
15
P( X = 13) = 0.613 0. 4 2 = 0.02194
13
15
P( X = 14 ) = 0.614 0. 41 = 0.00470
14
15
P( X = 15) = 0.615 0. 4 0 = 0.00047
15

For a 5% level of significance, what is the critical region?

Hence, under H0, the probability of

44

15 patients surviving

= 0.00047

14 or 15 patients surviving

= 0.00517

13, 14 or 15 patients surviving

= 0.02711

12, 13, 14 or 15 patients surviving

= 0.09050

Chapter 3 Hypothesis Testing: One Sample Tests

Note that if 14 surviving is adopted as a 'significant' result, the

probability of 15 must be included as well (because 15 is actually a
'better' result than 14). Similarly with 13, the probabilities for 14
or 15 must be included, and so on.
From the above (cumulative) probabilities it can be seen that the
critical region for say a 5% test is X 13
since

P( X 13) = 0.02711 < 0.05

but

P( X 12 ) = 0.09050 > 0.05

The number of patients who actually survived was 12. This value
does not lie in the critical region. Thus there is no evidence, at the
5% level of significance, to suggest that the drug is effective.
Note that rather than actually determine the critical region, it is
equally valid to simply argue that since
P( X 12 ) = 0.09050 > 0.05 , then H0 cannot be rejected.
Remember also that tables of the Cumulative Binomial
Distribution Function may often provide an easier and quicker
alternative for calculating probabilities or finding critical regions.

Activity 3

Can you tell the difference?

Can you tell HP Baked Beans from a supermarket brand? Can you
tell Coca Cola from a supermarket brand? Can you tell Stork from
butter?
You are going to set up an experiment to determine whether people
really can tell the difference between two similar foods or drinks.
Each person taking part in the test is given 3 samples; 2 of one
product and 1 of another (so they may have 2 beakers containing
Coca Cola (say) and 1 beaker containing a supermarket brand or
vice-versa).
Ask each person to try to identify the sample which is different
from the other two.
Note that there are 6 possible groupings of the 3 samples,
ABB

BAB

BBA

BAA

ABA

AAB

and a die may be used to decide which grouping to give each

individual person taking part.
Plan the experiment carefully before you start. Write out a list
showing the groupings for each of your participants (about 12,
say).

45

Chapter 3 Hypothesis Testing: One Sample Tests

Ensure that your participants take the test individually in quiet

surroundings, free from odours. All 3 samples must be of the
same size and temperature. If there are any differences in colour
then you can blindfold your participants. Record each person's
answer. Count the number of persons giving the correct answer.
Participants who are unable to detect any difference at all in the
3 samples must be left out of the analysis.
Use your results to test the hypotheses
H 0: p =
H 1: p >

1
3
1
3

(participants are simply guessing)

(participants can distinguish)

Under what conditions does a Poisson distribution provide a

suitable approximation to a binomial distribution?

Example
A machine which manufactures black polythene dustbin bags, is
known to produce 3% defective bags. Following a major
breakdown of the machine, extensive repair work is carried out
which may result in a change in the percentage of defective bags
produced. To investigate this possibility, a random sample of
200 bags is taken from the machine's production and a count
reveals 12 defective bags. What may be concluded?
Solution
Here n = 200, p = population proportion of defective bags
produced, and it is required to test
H0: p = 0.03

(no change)

H1: p 0.03

(change; two-tailed)

Significance level, = 0.05 (say)

Again from Chapter 6 of Statistics, a Poisson distribution with
= np provides an approximation to a binomial distribution
when n 50 and p 0.1 (or p 0.9 ).
Thus if X denotes the number of defective bags in the sample,
then under H0
X ~ Po( = 200 0.03 = 6.0 )

46

Chapter 3 Hypothesis Testing: One Sample Tests

Using tables of the Cumulative Poisson Distribution Functions,

P( X 12 ) = 1 P( X 11)
= 1 0.9799 = 0.0201
Since the test is two-tailed, this probability is compared with

= 0.025 , because a small number of defective bags could also

2
lead to the rejection of the null hypothesis.
Here, 0.0201 < 0.025 so H0 is rejected. There is evidence, at the
5% level of significance, that the percentage of defective bags
produced has changed following the repair work.
Under what conditions does a normal distribution provide a
suitable approximation to a binomial distribution?

Example
Company A proposes the take-over of Company B. The latter's
Chief Executive claims that her Company's shareholders are
equally divided for and against the take-over on the basis of the
terms offered. However, the Chairman of Company A claims that
more than half of Company B's shareholders are in favour of
accepting his Company's offer.
To investigate these two rival claims, the view of each of 400
randomly selected shareholders of Company B is sought. A
subsequent count reveals that 219 are in favour of the offer; the
remainder are against.
Does this provide evidence, at the 1% significance level, that the
claim made by Company B's Chairman is valid?
Solution
Let p denote the actual proportion of Company B's shareholders
who are in favour of the offer. Then
H0: p = 0.50 (claim of Company B's Chief Executive)
H1: p > 0.50 (claim of Company A's Chairman; one-tailed)
Significance level, = 0.01
If X denotes the number of Company B's shareholders in the
sample who are in favour of the offer, then under H0
X ~ B( 400, 0.50 )

with mean, np = 200 and

variance, np (1 p ) = 100
47

Chapter 3 Hypothesis Testing: One Sample Tests

From Chapter 8 of Statistics, a normal distribution provides an

approximation to a binomial distribution when n 50 and
0.1 p 0.9 .
Thus under H 0 , X ~ N (200,100 ) , or denoting the sample
X
then
proportion by P =
n

200 100
,
P ~ N
400 400 2
i.e.

P ~ N ( 0.50, 0.000625)

So

Critical region is z > 2.326

H00
Accept H

0.01
1%

2.326
Critical region
reject H00

Now value of sample proportion,

p =

219
= 0.5475
400

so test statistic,
z=

0.5475 0.50
= 1.90
0.000625

This value does not lie in the critical region. Thus there is no
evidence, at the 1% level of significance, to support the claim of
Company A's Chairman.

Exercise 3C
1. A pharmacist claims that more than 60% of all
customers simply collect a prescription. One of
her assistants notes that, in a random sample of
12 customers, 10 simply collected a prescription.
Does this provide sufficient evidence, at the 5%
level, to support the pharmacist's claim?

4. Tins of baked beans are packed in boxes of 24.

Results from a random sample of 25 boxes
delivered to supermarkets show that a total of 8
tins were damaged. Assess the claim that less
than 2% of tins are damaged during delivery.

2. In a survey carried out in Funville, 14 children

out of a random sample of 30 said that they
bought the Bopper comic regularly. Test, at the
10% level of significance, the hypothesis that the
true proportion of all children who buy this
comic regularly is 0.35.
(AEB)

5. A survey of Aldervale revealed that 223 houses

out of a random sample of 500 were fitted with
some form of double glazing. Test the
hypothesis that 40% of all houses in Aldervale
have some form of double glazing.
(AEB)

3. A random sample of 30 coffee drinkers were each

asked to taste-test a new brand of coffee. The
responses are listed below with L representing
'like', I representing 'indifferent', and D
representing 'dislike'.
L D L L D L L L L I
L L L I L D L I L L
I L L L D L L L L I

Do these data support the claim that more than

half of all coffee drinkers like this new brand of
coffee?

48

(AEB)

6. Prior to joining a coaching course, a netball

player has scored with 45% of her shots at the
basket, and so is considered to have a scoring
ability of 0.45. In games following the course,
she scores with 72 of her 135 shots at the basket.
Stating clearly any assumptions made,
investigate whether or not the course has
improved her shooting ability.

Chapter 3 Hypothesis Testing: One Sample Tests

3.4

Poisson population mean

In the previous Section it was seen how a Poisson distribution may

be used as an approximation to a binomial situation when testing
hypotheses concerning a population proportion. However, there
are situations for which a Poisson distribution is appropriate in its
own right when testing hypotheses concerning a population mean
value.
Under what conditions would you expect a Poisson distribution
to provide a suitable model for a random variable?
Recall from Chapter 6 of Statistics, that if X ~ Po( ) then
P( X = x ) =

xe
x!

x = 0,1,2,3, K,

with mean = and variance = .

What do you know about the sum of two independent Poisson
variables?

Example
The number of faults in one metre of a particular type of thread is
known to have a Poisson distribution. It is claimed that the
average number of faults is 0.02 per metre. A random sample of
100 one metre lengths of the thread reveals a total of 6 faults.
Does this information support the claim?
Solution
Remembering that if A ~ Po( a ) independent of B ~ Po( b ) then
C = A + B ~ Po( a + b ) , and denoting the total number of faults in
the 100 one metre lengths by X, then if the claim is valid
X ~ Po(100 0.02 = 2 )

It is thus required to test

H 0: = 2
H 1: 2

(two-tailed)

Significance level, = 0.05 (say)

The observed number of faults is 6, and using tables of the
Cumulative Poisson Distribution Function with = 2
P ( X 6 ) = 1 P ( X 5)
= 1 0.9834 = 0.0166 < 0.025 (two-tailed)

49

Chapter 3 Hypothesis Testing: One Sample Tests

There is evidence, at the 5% level of significance, that the claim is

not valid.

Activity 4

Simulated traffic flow

The random variable X denotes the number of vehicles in a

30
second interval passing a traffic check point on a country road.
Use a three-digit random number, y, to generate an 'observed'
value, x, of X using the following rules.
000 y 548

x=0

549 y 877
878 y 976
977 y 996
997 y 999

x =1
x=2
x=3
x=4

Using the above rules, obtain 5 'observed' values of X.

Use the total of these values to test the hypothesis that the average
number of vehicles passing the traffic check point in 30 seconds is
0.60.
Repeat the test for 10, 15 and 20 simulated 'observed' values of X.
Consider the conclusions of your four tests.

For a Poisson distribution with a mean greater than 15, a normal

distribution with = 2 = provides a good approximation (see
Chapter 8 of Statistics).

Example
Prior to extensive modernisation of its forecourt, the average
number of vehicles calling for fuel at a small garage has been 5 per
hour. Following the modernisation, a total of 543 vehicles called
for fuel in a random sample of 100 one-hour intervals. Has the
modernisation of the garage's forecourt significantly increased the
average number of vehicles calling for fuel?
Solution
If X denotes the total number of vehicles calling for fuel in the 100
one-hour intervals, then under the null hypothesis of no increase
X ~ Po(100 5 = 500 )

50

N (500, 500 )

Chapter 3 Hypothesis Testing: One Sample Tests

The test is thus as follows.

H0: = 500
H1: > 500

(one-tailed)

5%
0.05

H00
Accept H

Significance level, = 0.05 (say)

Critical region is z > 1.645

1.645
Critical region
H00
reject H

543 500
= 1.92
Test statistic, z =
500

This value does lie in the critical region. Thus there is

evidence, at the 5% level of significance, to suggest that the
forecourt modernisation has resulted in an increase in the
average number of vehicles calling for fuel.

Exercise 3D
1. In the manufacture of commercial carpet on a
particular machine, small faults occur at random
in the carpet at an average rate of 0.925 per
25 m 2 . Following an overhaul of the machine, an
inspection of a random sample of four 5 m 5 m
squares of carpet reveals only 2 small faults. Is
there evidence, at the 5% level of significance,
that, following the overhaul, the mean number of
small faults has been reduced?

4. It has been established over a period of time that

the weekly numbers of breakages of crockery and
glassware in a hotel follow a Poisson distribution
with mean 27. On appointment, a new manager
announces to staff that unless this mean level is
reduced, bonuses will suffer as a result.
Following this announcement, the weekly
numbers of breakages over an eight-week period
were as follows.

2. It is known that the numbers of parasites on fish

in a pond follow a Poisson distribution. The
total number of parasites on a random sample of
3 fish taken from the pond is 8. Test the
hypothesis that the mean number of parasites per
fish is 4.
(AEB)

Do these data provide evidence that the

managers announcement has been heeded by the
staff?

23

19

17

20

25

32

22

26

3. The daily number of letters of complaint from

customers received by a department store follows
a Poisson distribution. Over a 150-day period, a
total of 407 letters of complaint were received.
Investigate, at the 1% level of significance, the
claim that an average of fewer than 3 letters of
complaint per day are received. What
assumption is needed regarding the
150-day period?
(AEB)

3.5

Population median

In Chapter 3 of Statistics, the median, m, of n items of data is

n + 1
defined as the
th item of data; or, in other words, the
2
middle item.
51

Chapter 3 Hypothesis Testing: One Sample Tests

For a random variable, X, with median , this leads to the result

P( X < ) = P( X > ) = 0.5
or

P( X < 0 ) = P( X > 0 ) = 0.5

or

P( X is negative ) = P( X is positive ) = 0.5

These equally probable positive and negative values lead to a

test for a hypothesised value of a population median based upon
a binomial distribution with p = P( + sign ) = 0.5 under H0.
Note however that, contrary to all the other tests in this chapter,
no assumption needs to be made about the distribution of X
itself. As a result this sign test for a population median is an
example of a non-parametric test.
Tests which require a knowledge of, or assumption about, the
distribution of the parent population (e.g. normal, binomial,
Poisson) are called parametric tests.

Example
The values below are the scores (maximum 20) obtained in an
aptitude test by a random sample of 11 graduates. It is known
that for the non-graduate population the median score is 12. Is
there evidence, at the 10% significance level, that graduates
achieve a higher median score than the non-graduate population?
14 15

9 10 10 13 14 19 12 16 13

Solution
H0: = 12
H1: > 12

(one-tailed)

Significance level, = 0.10

Signs of (score 12) are

+ +

+ 0 +

Let X denote the number of + signs. Then, ignoring the one 0

in this case, under H0,
X ~ B (10, 0.50 ) with observed value of X = 7 .
Now

P ( X 7) = 1 P ( X 6 ) = 1 0.8281 = 0.1719 > 0.10

Thus there is no evidence, at the 10% level of significance, to

suggest that graduates achieve a higher median score than the
non-graduate population.
52

Chapter 3 Hypothesis Testing: One Sample Tests

What additional information would be required in the above

example in order to test H 0 : = 12 versus H1: > 12 ?

Activity 5

Weights and measures re-visited

Using your data collected for Activity 1, test the hypothesis that
the true median value for the contents is equal to the value
stated as contents.
Explain the reasons for the similarity, or difference, between
your conclusions for the two activities.

Exercise 3E
1. The following values are the annual salaries,
in , of a random sample of 10 recent statistics'
graduates.
13 250
7 485 15 136 12 258 11 019
14 268 19 536 14 326 16 326 17 984

Investigate the claim, at the 10% significance

level, that the median annual salary of all recent
statistics' graduates exceeds 12 000.
2. Part of the assessment of a statistics course
involves small groups of students giving a short
presentation of their findings. The grades
obtained by 13 such groups are as follows
(highest = A, lowest = F).

4. During a working day a machine may require

various adjustments. The machine's operator is
asked to record the number of adjustments made
to the machine each day for a period of 250
working days. She obtains the data displayed in
the table below.
Number of
adjustments
Number of days

5 6

61 54 65 18 12

9 31

Test the hypothesis that the true median daily

number of adjustments is 1.
Explain why a test for the true mean daily
number of adjustments would be difficult.

C B A B B C D C F B A B B

Test the hypothesis that the median grade for all

such group presentations is C.
3. It is claimed by a local resident that more than
50% of all the vehicles on an urban road exceed
the 30 mph speed limit. The speed of each of a
random sample of 24 vehicles is recorded with
the following results.
42 35 24
34 30 29
34 39 43

30 32 42 56 35
41 38 38 30 29
72 38 40 30 62

Investigate the resident's claim.

53

Chapter 3 Hypothesis Testing: One Sample Tests

3.6

Miscellaneous Exercises

1. An investigation was conducted into the dust

content in the flue gases of a particular type of
solid-fuel boiler. Thirteen boilers were used
under identical fuelling and extraction
conditions. Over a similar period, the following
quantities, in grams, of dust were deposited in
traps inserted in each of the thirteen flues.
73.1

56.4

82.1

67.2

78.7

75.1

53.3

55.5

61.5

60.6

55.2

63.1

48.0

Stating any necessary distributional assumptions,

test, at the 5% level of significance, the
hypothesis that
(a) the population variance of dust deposit is
35 g 2 ,
(b) the population mean dust deposit is 60 g.
(AEB)
2. As part of a research project, a random sample of
11 students sat a proposed national Physics
examination and obtained the following
percentage marks.
30 44 49 50 63 38 43 36 54 40 26

(a) Use a t-test, at the 10% significance level, to

test the hypothesis that the true mean
examination mark is 40%.
(b) Use a sign test, at the 10% level, to test the
hypothesis that the median mark for all
candidates is 40%.
What assumption is needed for (a) but not for (b),
and do you think it is justified?
(AEB)
3. Smallwoods Ltd run a weekly football pools
competition. One part of this involves a fixedodds contest where the entrant has to forecast
correctly the result of each of five given matches.
In the event of a fully correct forecast the entrant
is paid out at odds of 100 to 1. During the last
two years Miss Fortune has entered this fixedodds contest 80 times. The table below
summarises her results.
Number of matches
correctly forecast
per entry (x)

Number of entries with

x correct forecasts (f)

4 13 27 22 13

Assuming that the number of matches correctly

forecast per entry follows a binomial
distribution, test the hypothesis that the
probability of Miss Fortune forecasting correctly
the result of a single match is 0.5.
On the evidence before you, and assuming that
the point of entering is to win money, would you
advise Miss Fortune to continue with this
competition, and why?
(AEB)

54

4. At a nuclear power station great care is taken to

monitor employees' state of health. The table
below gives the number of visits made by each
of 10 employees from the reactor room to their
general practitioners during one calendar year.
3

The number of such visits made by a member of

the general public is known to have a Poisson
distribution with mean 3. Assuming that the
numbers of visits made by the nuclear power
station's employees are also distributed as a
Poisson random variable, test the hypothesis that
the annual mean per employee is greater than 3.
(AEB)
5. A random sample of 18 female computer
operators each had their diastolic blood pressure
measured to the nearest millimetre with the
following results.
57

64

77

82

66

94

72

83

61

72

65

89

54

73

55

67

71

68

(a) Stating any necessary distributional

assumptions, show that there is no significant
evidence to reject the hypothesis that the
standard deviation of female computer
operators' diastolic blood pressures is equal
to that of the female population, namely
10 mm.
(b) Making appropriate use of the statement in
(a), investigate the claim that the mean
diastolic blood pressure of female computer
operators differs from 75 mm.
(AEB)
6. The number of misprints per page in each
national daily newspaper is known to follow a
Poisson distribution. The mean number per page
for the tabloid newspapers is 4. The Daily
Planet, a quality newspaper, claims that although
its pages are much bigger and contain more text
than those of the tabloid papers, its mean number
of misprints per page is certainly fewer.
A random sample of 16 pages from recent
editions of the Daily Planet results in the
following numbers of misprints.
0, 3, 2, 3, 1, 4, 5, 2, 3, 4, 5, 2, 4, 3, 3, 1.

Investigate, at the 1% significance level, the

Daily Planet's claim.
(AEB)
7. Explain what is meant by the following terms
when used in the context of a hypothesis test.
(a) Null and alternative hypotheses.
(b) Type 1 and Type 2 errors.
(c) One-tailed and two-tailed tests.
A new dietary treatment for a severe allergy is
claimed to have a better cure rate than the
accepted value of 60% for the well-established
standard drug treatment. A random sample of 20
patients, suffering from the allergy, are given the
new dietary treatment and as a result 17 are
cured. Is the claim valid?
(AEB)

Chapter 3 Hypothesis Testing: One Sample Tests

8. A large food processing firm was considering

introducing a new recipe for its ice cream. In a
preliminary trial, a panel of 15 tasters were each
asked to score the ice cream on a scale from 0
(awful) to 20 (excellent).Their scores were as
follows.
16

15

17

18

15

18

12

14

19

14

17

The mean score in a similar trial for the firm's

existing ice cream was 14. Test the hypothesis
that the mean score for the new ice cream was no
different.
Because of the erratic nature of the scores
obtained, doubt was expressed as to the
assumption that scores were normally
distributed. As a result it was suggested that a
sign test for a median was perhaps more
appropriate. Given that the median score in the
trial for their existing ice cream was 13, what
can be concluded?
(AEB)
9. The external diameter, in centimetres, of each of
a random sample of 10 pistons manufactured on
a particular machine was measured with the
results below.
9.91,

9.89, 10.06,

9.81, 10.01,

9.99,

9.98, 10.09,
9.87, 10.09.

Stating any necessary assumptions, test the two

distinct claims that piston rings manufactured on
this machine have a mean external diameter of
(AEB)
10 cm with a variance of 0.005 cm 2.
10. A sweet shop sells chocolates which appear, at
first sight, to be identical. Of a random sample
of 80 chocolates, 61 had hard centres and the rest
soft centres. Test the hypothesis that 70% of
chocolates have hard centres.
The chocolates are all in the shape of circular
disks and the diameters, in millimetres, of the 19
soft centred chocolates were as follows.
279

263

284

277

281

269

271

262

275

266

272

281

274

279

277

267

269

275

266

Assuming that the diameters of the soft centred

chocolates are normally distributed, test, at the
10% significance level, the hypothesis that their
mean diameter is 275 mm.
What changes would you make to your test if it
was known that the standard deviation of the
diameters of soft centred chocolates was 5 mm?
(AEB)
11. It is known that repeated weighings of the same
object on a particular chemical balance give
readings which are normally distributed. Past
evidence, using experienced operators, suggests
that the mean is equal to the mass of the object
and that the standard deviation is 0.25 mg.

A trainee operator makes seven repeated

weighings of the same object, which is known to
have a mass of 19.5 mg, and obtains the
following readings.
19.1 19.4 19.0 18.8 19.7 19.8 19.3

Is there any evidence that these results are more

variable than those obtained by experienced
operators?
Hence investigate whether or not the trainee
operator's readings are biased.
(AEB)
12. As part of a statistics project, students observed
five private cars passing a college and counted
the number which were carrying the driver only,
with no passengers. This was repeated 80 times.
The results for a particular student were as
follows.
Number of cars
with driver only

Number of times
observed

12

27

26

12

Explain why the distribution of the number of

cars with a driver only could be modelled by a
binomial random variable. Assuming such a
model is appropriate, state the value of the
parameter n and estimate the value of the
parameter p.
Hence investigate the claim that more than 60%
of cars contain the driver only.
(AEB)
13. Explain each of the terms significance level,
critical region and test statistic as used in
hypothesis testing.
The manager of a road haulage firm records the
following times taken, in minutes, by a lorry to
travel from the depot to a particular customer's
factory.
43

35

47

180

39

58

40

39

51

The journey time of 3 hours was as a result of the

driver being stopped by Customs & Excise
Inspectors. The manager therefore removes this
value before passing the data to you, as the firm's
statistician, for analysis. Use the eight remaining
values to test the hypothesis that the true mean
journey time is 40 minutes.
Comment on the manager's decision to remove
the value of 3 hours and state what assumption
may have been violated if this value had been
included.
What alternative test would you have considered
if the manager had requested that all nine results
should be analysed?
(AEB)

55

Chapter 3 Hypothesis Testing: One Sample Tests

14. A car insurance company found that the average

amount it was paying on bodywork claims in
1992 was 435 with a standard deviation of 141.
The first six bodywork payments, in , in 1993
were
548

209

534

198

789

633.

15. (a) Before its annual overhaul, the mean

operating time of an automatic machine was
100 seconds. After the overhaul, the
following random sample of operating times,
in seconds, was obtained.
97 101 92 101 95

95

98

96

95

Assuming that the time taken by the machine

to perform the operation is a normally
distributed random variable, test the
hypothesis that the overhaul has improved the
machine's operation.
(b) The results of a survey showed that 3615 out
of 10 000 families regularly purchased a
specific weekly magazine. Test, at the 1%
level of significance, the claim that more than
35% of families buy the magazine.
(AEB)
16. A nurseryman decided to keep records of the first
year's growth of his pine seedlings. On the first
occasion he found a mean growth of 11.5 cm
with a standard deviation of 2.5 cm. The
following year he used an experimental soil
preparation for all his seedlings and the first
year's growth of a random sample of eight of the
seedlings was
7, 23, 19, 25, 11, 18, 17 and 15 cm.

(a) Assuming these data may be regarded as a

random sample from a normal distribution,
test at the 5% significance level, whether
there has been a change in
(i) the standard deviation,
(ii) the mean.
(b) Explain to the nurseryman why the
conclusions in (a) cannot necessarily be
attributed to the new soil preparation. (AEB)
17. The development engineer of a company making
razors records the time it takes him to shave, on
seven mornings, using a standard razor made by
the company. The times, in seconds, were
217, 210, 254, 237, 232, 228, 243.

Assuming that this may be regarded as a random

sample from a normal distribution with mean
and variance 2 test, at the 5% level of
significance, the hypothesis that
(a) = 10

seconds,

(b) = 240 seconds.

56

218 207 214 189 211 206 203 217 183 186
219 213 207 214 203 204 195 197 213 212

Stating clearly all necessary assumptions, has

there been a significant change in the average
and in the variability of payments in 1993 as
compared to 1992?
(AEB)

90

18. Packets of ground filter coffee have a nominal

weight of 200 g. The distribution of weights
may be assumed to be normal. A random sample
of 30 packets had the following weights.

(AEB)

188 221 217 184 186 216 198 211 216 200

Investigate the assumption that the mean weight

of all packets is 200 g.
Test the hypothesis that 15% of packets weigh
less than 190 g.
(AEB)
19. (a) The number of accidents per day on a stretch
of motorway is known to have a Poisson
distribution. Police claim that there is an
average of more than one accident per day.
Over a particular 7-day period there is a total
of 13 accidents. Investigate the claim of the
police.
(b) The results of a survey to establish the
attitude of individuals to a particular
proposal showed that three quarters of those
interviewed were house owners. Of the 200
interviewed, only 12 of the 70 in favour of
the proposal were not house owners. Test the
hypothesis that the percentage of house
holders in favour of the proposal is 40.
(AEB)
20. The resistances (in ohms) of a sample from a
batch of resistors were as follows.
2314

2456

2389

2361

2360

2332

2402

Stating any necessary assumptions, test the

hypothesis that the true standard deviation of the
resistors is 30 ohms.
Hence test the hypothesis that the actual mean
resistance of the resistors is 2400 ohms. (AEB)
21. Explain the difference between parametric and
non-parametric hypothesis tests, and give an
example of each type.
A local authority offers all of its employees
regular health checks. As part of the check,
several physiological measurements are taken on
each person.
(a) The ordered scores for one of the
measurements on nine employees were as
follows.
9 11 50 54 58 69 76 91 95

Test the hypothesis that the median score for

this measurement is 50.
(AEB)
(b) For another measurement, a random sample
of 100 employees had 19 scores of exactly 50
and 30 scores above 50. Investigate the
claim that the median score for this
measurement is less than 50.
(AEB)

Chapter 3 Hypothesis Testing: One Sample Tests

22. Experimental components for use in aircraft

engines were tested to destruction under extreme
conditions. The survival times, X days, of ten
components were as follows.
207 381 411 673 534 294 697 344 418 554

(a) Assuming that the survival time, under these

conditions, for all the experimental
components is normally distributed, assess
the validity of the claim that the mean
survival time exceeds 400 days.
(b) Given that the standard deviation of survival
times is known to be 150 days, what changes
would you make to your test procedure
in (a)?
(AEB)
23. The external diameters (measured in units of
0.01 mm above nominal value) of a sample of
piston rings produced on the same machine were
11

32

18

29

11

21

19

24. Explain each of the terms null hypothesis,

critical region and test statistic as used in
hypothesis testing.
Employees of a firm carrying out motorway
maintenance are issued with brightly coloured
waterproof jackets. These come in different sizes
numbered 1 to 5. The last 40 jackets issued were
of the following sizes.
2
5
2
1

3
4
4
3

3
1
4
4

1
2
1
3

3
3
5
3

3
3
3
2

2
2
3
5

4
4
2
1

3
5
3
4

2
3
3
4

Assuming that the 40 employees may be regarded

as a random sample of all employees, test the
hypothesis, at the 5% significance level, that
40% of all employees require size 3.
Test the claim that size 3 is the median size.
(AEB)

(a) Assuming a normal distribution, and given

that the target value is 20 and a standard
deviation of at most 10 is acceptable, carry
out appropriate hypothesis tests so as to
comment on the performance of the machine.
(b) It was later discovered that an error had been
made in zeroing the measuring device and
that all the measurements in the sample
should be increased by 12. How, if at all,
does this affect your conclusions in (a)?
(AEB)

57

Chapter 3 Hypothesis Testing: One Sample Tests

58