Topic 9 QBA241 2015

9. Hypothesis Testing

This topic will present the general procedure of making statistical inferences with regards to the
population using a method known as ‘hypothesis testing’.

The objective of this procedure is to examine the sample data and determine if there is a sufficient
evidence to support the hypothesis.

9.1 The Concepts of Hypothesis Testing

A hypothesis is a statement or a claim made concerning the value of the population parameter.
For examples:

 The mean income of employees at the McDonald’s is AED3,000 per month

 The mean time it takes to finish writing a Statistics exam at least 1.5 hours
 The percentage of adults who own a smart phone is greater than 80%
 The proportion of children who spend watching television less than 5 hours daily is not
more than 50%

There are two hypotheses; null hypothesis and alternative hypothesis. A null hypothesis is
represented as ‘H0’ while the alternative hypothesis is represented as ‘H1’. It is always about the
population parameter such as “µ” or “p”. Never about the sample statistic.

A null hypothesis always states about the ‘status quo’ of the population parameter. For example:

H0: µ = AED3,000
(The mean income of employees at the McDonald’s is AED3,000 per month)

An alternative hypothesis always challenges the ‘status quo’ of the population parameter. For
example:
H1: µ ≠ AED3,000
(The mean income of employees at the McDonald’s is not AED3,000 per month)

The general rule of thumb for stating the null hypothesis (H0):
 It must contain “=” or “≤” or “≥”

The general rule of thumb for stating the alternative hypothesis (H1):
 It must never contain “=” or “≤” or “≥”
 Hence, it always consists of “≠” or “>” or “<”

Hypothesis Testing 73 HH
Topic 9 QBA241 2015

The purpose of testing the hypothesis is to conclude either to reject or unable to reject the null
hypothesis. If the null hypothesis is rejected, it indicates that there is sufficient evidence to support
the null hypothesis. However, if we fail to reject the null hypothesis, it indicates that there is
insufficient evidence to support the null hypothesis.

There are two possible errors in hypothesis testing. A Type I error is committed when a true null
hypothesis is rejected. A Type II error is committed when a false null hypothesis is not rejected.

The probability of a Type I error is also known as the level of significance (α). The probability of a
Type II error is denoted by β (“beta”). Both error probabilities are inversely related which means in
an attempt to reduce the α, the β will increase and vice-versa.

The concepts of hypothesis testing are best known to the analogy used in a criminal trial. When a
person is accused of a crime, he or she faces a trial. The prosecution presents the case and the jury
must make a decision on the basis of the evidence presented. If there is sufficient evidence to
incriminate the defendant, then he or she will be convicted and punished. On the other hand, if
there is insufficient evidence to incriminate the defendant, then he or she will walk free.

Therefore, by default, the null hypothesis should state the status quo whereas the alternative
hypothesis challenges it.

H0: The defendant is innocent. H1: The defendant is guilty.

The Type I error is committed when the defendant is convicted when he or she is in fact innocent.
Then the Type II error is committed when the defendant is acquitted when he or she is in fact guilty
of the crime. The analogy of errors in a criminal trial is summarized in the table below:

Table 1: Analogy of a Criminal Trial

DECISION H0 IS TRUE H0 IS FALSE
Do Not Reject H0 Confidence Interval = (1 – α) Type II error (β)
Reject H0 Type I error (α) Power of the Test = (1 – β)

9.1.1 Steps in Testing a Hypothesis

The steps of testing a hypothesis include the following:

1) Establish the null and alternative hypotheses
2) Set the level of significance ()
3) Determine the critical value and its rejection region
4) State the decision rule
5) Compute the standardized test statistic by converting the sample statistic
6) Draw conclusion by comparing the test-statistic against the critical value

Hypothesis Testing 74 HH
Topic 9 QBA241 2015

To illustrate the steps in testing a hypothesis, let’s refer to the following example.

A manufacturer of chocolate bars claims that the mean weight is 3 ounces with a standard deviation
of 0.8 ounces. A quality inspector insists on examining the manufacturer’s claim by collecting a
sample of 100 chocolate bars and weight them. The mean weight of chocolate bars from the sample
is 2.84 ounces. Is there sufficient evidence to infer that the manufacturer’s claim is true? Use α =
0.05.

Solution

Step 1: Hypothesis
H0: µ = 3 ounces (The mean weight of chocolate bars is 3 ounces.)
H1: µ ≠ 3 ounces (The mean weight of chocolate bars is not 3 ounces.)

This type of hypothesis is known as the two-tailed hypothesis.

Step 2: Level of Significance

α = 0.05

Step 3: Critical Value

Zα/2 = Z0.05/2 = ±1.96

Step 4: Decision Rule

Reject H0 if the test-statistic is less than Z0.05/2 = 1.96 or greater than Z0.05/2 = 1.96. Otherwise, do
not reject H0.

Step 5: Test-Statistic
Xμ 2.84  3  0.16
ZSTAT     2.0
σ 0.8 0.08
n 100

Step 6: Conclusion
Since the ZSTAT = 2.0 is less than 1.96, the null hypothesis is rejected. There is sufficient evidence
to infer that the mean weight of chocolate bars is not 3 ounces. Hence, the manufacturer’s claim is
not substantiated.

Hypothesis Testing 75 HH
Topic 9 QBA241 2015

9.1.2 Level of Significance, Critical Value and Rejection Region

Recall that the level of significance is denoted as “α”. It is also known as the probability of
committing a Type I error.

The level of significance is usually stated by the researcher. The common significance levels are
1%, 5% and 10%. (If the significance level is not given, then the default value is 5%)

The critical value is then determined using the information from the hypothesis and significance
level. The value is referred to the Standardized Normal Distribution Table (z-table) or the Student’s
t-distribution (t-table).

The rejection region is the region where the null hypothesis is rejected. It is determined based on
the direction of H1. The size of rejection region is based on the significance level.

Refer to the following example. Suppose that there’s a claim the population mean age is 50 years
old.

Hypothesis Testing 76 HH
Topic 9 QBA241 2015

Sample of an Upper-Tailed Test Sample of a Lower-Tailed Test

 

Critical Value Critical Value

Reject H0 if the test statistic > critical value Reject H0 if the test statistic < critical value

9.2 Testing the Hypothesis of a Population Mean

When testing a hypothesis involving the population mean (µ), it is imperative to consider if the
value of the population standard deviation () is known. The rule of thumb is:

 If  is known, the critical value of the Standardized Normal Distribution (z-table) is used,
i.e. Zα or Zα/2
 If  is unknown, the critical value of the Student’s t-Distribution (t-table) is used, i.e. tα,n-1 or
tα/2,n-1

9.2.1 Testing the Population Mean ( Known)

Example 1: Upper-Tailed Test

Suppose that the manager of a luxury brand boutique is considering in establishing a special
discount system for its loyal customers. After a thorough financial analysis, he determine that the
special discount system will be cost-effective only if the mean amount spent is greater than
AED15,000 per month. A random sample of 50 accounts is selected which reveals a sample mean of
AED16,800. The manager knows that accounts are normally distributed with a standard deviation
of AED7,570. Is there sufficient evidence to infer that the special discount system will be cost-
effective?

Solution

Step 1: Hypothesis
H0: µ ≤ AED15,000 (The mean amount spent is not greater than AED15,000)
H1: µ > AED15,000 (The mean amount spent is greater than AED15,000)

This type of hypothesis is known as the upper-tailed hypothesis.

Hypothesis Testing 77 HH
Topic 9 QBA241 2015

Step 2: Level of Significance

α = 0.05 (taken by default)

Step 3: Critical Value

Zα = Z0.05 = 1.645

Step 4: Decision Rule

Reject H0 if the test-statistic is greater than Z0.05 = 1.645. Otherwise, do not reject H0.

Rejection Region

Non
Rejection α = 0.05
Region

Zα = Z0.05 = 1.645
Step 5: Test-Statistic

𝑥̅ − 𝜇 16800 − 15000
𝑍𝑆𝑇𝐴𝑇 = 𝜎 = = 1.681
7570
√𝑛 √50

Step 6: Conclusion
Since the ZSTAT = 1.681 is greater than Z0.05 = 1.645, the null hypothesis is rejected. There is sufficient
evidence to infer that the mean amount spent is greater than AED15,000 per month. Based on this
finding, it indicates that the special discount system will be cost-effective.

Example 2: Lower-Tailed Test

Karina claims that she can finish a 100 meter sprint under 12 seconds with a standard deviation of
1.35 seconds. Karina’s trainer would like to investigate her claim. He records a sample of 30 runs
for the past month which reveals an average of 11.7 seconds. Is there sufficient evidence to support
Karina’s claim? Use α = 0.10.

Solution

Step 1: Hypothesis
H0: µ ≥ 12 seconds (The mean sprint time is at least 12 seconds)
H1: µ < 12 seconds (The mean sprint time is less than 12 seconds)

This type of hypothesis is known as the lower-tailed hypothesis.

Step 2: Level of Significance

α = 0.10

Hypothesis Testing 78 HH
Topic 9 QBA241 2015

Step 3: Critical Value

Zα = Z0.10 = 1.28

Step 4: Decision Rule

Reject H0 if the test-statistic is less than Z0.10 = 1.28. Otherwise, do not reject H0.

Rejection Region
Non

Rejection
Region

Zα = Z0.10 = 1.28

Step 5: Test-Statistic
𝑥̅ − 𝜇 11.7 − 12
𝑍𝑆𝑇𝐴𝑇 = 𝜎 = = −1.217
1.35
√𝑛 √30

Step 6: Conclusion
Since the ZSTAT = 1.217 is greater than Z0.10 = 1.28, the null hypothesis is not rejected. There is
insufficient evidence to infer that the mean sprint time is less than 12 seconds.

9.2.2 Testing the Population Mean ( Unknown)

If the population standard deviation () is unknown then the sample standard deviation (s) will be
used as the point estimate. Instead of using the z-value as the critical value, the Student’s t-
Distribution (t-table) is used, i.e. tα,n-1 or tα/2,n-1.

Example 1: Two-Tailed Test

The mean cost of a 3-star hotel room in Niagara Falls is said to be $168 per night. A random sample
of 25 hotels resulted in 𝑥̅ = $172.50 and s = $15.40. Test at α = 0.05.

Solution

Step 1: Hypothesis
H0: µ = $168 (The mean cost is $168 per night)
H1: µ ≠ $168 (The mean cost is not $168 per night)

This type of hypothesis is known as the two-tailed hypothesis.

Step 2: Level of Significance

α = 0.05

Hypothesis Testing 79 HH
Topic 9 QBA241 2015

Step 3: Critical Value

tα/2,n-1 = t0.05/2,25-1 = ±2.064

Step 4: Decision Rule

Reject H0 if the test-statistic is less than t0.05/2,24 = 2.064 or greater than t0.05/2,24 = 2.064.
Otherwise, do not reject H0.

Step 5: Test-Statistic
𝑥̅ − 𝜇 172.50 − 168
𝑡𝑆𝑇𝐴𝑇 = 𝑠 = = 1.46
15.40
√𝑛 √25

Step 6: Conclusion
Since the tSTAT = 1.46 falls within the two critical values (t0.05/2,25-1 = ±2.064), the null hypothesis is
not rejected. There is insufficient evidence to infer that the mean cost of a 3-star hotel room in
Niagara Falls is $168 per night.

Hypothesis Testing 80 HH
Topic 9 QBA241 2015

9.2.3 Testing the Population Proportion (p)

When the emphasis is on the degree to which a population possesses a particular characteristic
(known as “success”), the measure of interest is a percentage or a proportion.
𝑥
̂𝑝 =
𝑛
where, x = number of successes (occurrences) in a sample
n = sample size

A proportion refers to the fraction of the total that possesses a certain characteristic. For example,
we might ask what proportion of women in our sample weigh less than 110 pounds. Since 2 women
out of 10 women weigh less than 110 pounds, the proportion would be x/n or 2/10 or 20%.

A percentage is another method of expressing a proportion. A percentage is equal to the proportion

multiply it by 100. In our example of the 10 women, the percent of the total who weigh less than
110 pounds would be 100 * (2/10) or 20 percent.

Example 1: Two-Tailed Test

A marketing company claims that it receives 8% responses from its mailing. To test this claim, a
random sample of 500 were surveyed with 30 responses. Test at α = 0.05.

Solution

Step 1: Hypothesis
H0: p = 0.08 (The response rate from mailing is 8%)
H1: p ≠ 0.08 (The response rate from mailing is not 8%)

This type of hypothesis is known as the two-tailed hypothesis.

Step 2: Level of Significance

α = 0.05

Step 3: Critical Value

Zα/2 = Z0.05/2 = ±1.96

Step 4: Decision Rule

Reject H0 if the test-statistic is less than Z0.05/2 = 1.96 or greater than Z0.05/2 = 1.96. Otherwise, do
not reject H0.

Hypothesis Testing 81 HH
Topic 9 QBA241 2015

Step 5: Test-Statistic
𝑥 30
̂𝑝 = = = 0.06
𝑛 500

𝑝̂ − 𝑝 0.06 − 0.08
𝑍𝑆𝑇𝐴𝑇 = = = −1.648
√𝑝(1 − 𝑝) √0.08(1 − 0.08)
𝑛 500

Step 6: Conclusion
Since the ZSTAT = 1.648 falls within the two critical values (Z0.05/2 = ±1.96), the null hypothesis is not
rejected. There is insufficient evidence to infer that the response rate from the mailing is 8% as per
claim.

Example 2: Lower-Tailed Test

The Human Resources Department of a large insurance company is interested in reducing the
turnover rate of data processing clerks in the first year of employment. Past records indicate that
25% of all new employees in this area are no longer employed at the end of the first year. Extensive
new training approaches are implemented for a sample of 150 new data processing clerks. At the
end of the first year 29 of these 150 new recruits were no longer employed. At the 1% significance
level, is there evidence to indicate that the proportion of data processing clerks who have gone
through the new training and are no longer employed is less than 25%?

Solution

Step 1: Hypothesis
H0: p ≥ 0.25
(The proportion of data processing clerks who have gone through the new training and are no
longer employed is not less than 25%)
H1: p < 0.25
(The proportion of data processing clerks who have gone through the new training and are no
longer employed is less than 25%)

This type of hypothesis is known as the lower-tailed hypothesis.

Step 2: Level of Significance

α = 0.01

Hypothesis Testing 82 HH
Topic 9 QBA241 2015

Step 3: Critical Value

Zα = Z0.01 = 2.33

Step 4: Decision Rule

Reject H0 if the test-statistic is less than Z0.01 = 2.33. Otherwise, do not reject H0.

Rejection Region


Non-Rejection
Region

Zα = Z0.01 = 2.33

Step 5: Test-Statistic
𝑥 29
̂𝑝 = = = 0.1933
𝑛 150

𝑝̂ − 𝑝 0.1933 − 0.25
𝑍𝑆𝑇𝐴𝑇 = = = −1.604
√𝑝(1 − 𝑝) √0.25(1 − 0.25)
𝑛 150

Step 6: Conclusion
Since the ZSTAT = 1.604 is greater than Z0.01 = 2.33, then the null hypothesis is not rejected. There
is insufficient evidence to infer the proportion of data processing clerks who have gone through the
new training and are no longer employed is less than 25%.

Hypothesis Testing 83 HH
Topic 9 QBA241 2015

EXERCISES

1. A standard intelligence examination has been given for several years with an average score of
80 and a standard deviation of 7. If 25 students taught with special emphasis on reading skill,
obtain a mean grade of 83 on the examination, is there reason to believe that the special
emphasis improves the result on the test? Use  = 0.05.

2. In a recent year, 73% of 1st-year college students responding to a national survey identified
“being very well-off financially” as an important personal goal. In a random sample of 200 of its
1st-year students, a state university finds that 132 students say that this goal is important. Test
the hypothesis that the proportion of 1st-year students at this university who think this goal is
important differs from the national value of 73%. Use  = 0.10.

3. The birth weight of babies is normally distribution with a mean of 3.8 kg with a standard
deviation of 0.80 kg. A researcher suspects that the weight of babies from mothers who smoked
a lot during pregnancy will be lower than the population mean. To examine this, he takes a
random sample of 26 babies from mothers who smoked a lot. The mean birth weight in this
group was 3.48 kg. Use  = 0.05.

4. A university has found over the years that out of all the students who are offered admission, the
proportion of students who accept the offer is 70%. After a new director of admissions is hired,
the university wants to check if the proportion of students accepting has increased significantly.
Suppose they offer admission to 1200 students and 888 students accept the offer. Is there
evidence at the 0.05 level of significance that there has been a significant increase in proportion
of students accepting admission?

5. Past experience shows that, if a certain machine is adjusted properly, 5 percent of the items
turned out by the machine are defective. Each day the first 25 items produced by the machine
are inspected for defects. One item is found to be defective. Is there sufficient evidence to infer
that the defective rate is lower than 5 percent? Use  = 0.05.

6. A nationwide survey of college students was conducted and found that students spent on
average two hours per class hour studying with a standard deviation of 0.24 hours. A professor
at your college wants to determine whether the time students spent at your college is
significantly different from two hours. A random sample of 15 statistics students is carried out
and the findings indicate an average of 1.75 hours. Determine whether the average time
students spent at your college is significantly different from two hours at the 5% significance
level.

7. The mean length of a small counterbalance bar is 45 mm. The production supervisor is
concerned that the adjustment of the machine producing the bars has changes. He asks the
engineering department to investigate. The engineering department selects a random sample of
12 bars which reveals a mean of 44 mm and a standard deviation of 1.784 mm. Is it reasonable
to conclude that there has been a change in the mean length of bars? Use  = 0.05.

Hypothesis Testing 84 HH