26134 Business Statistics Hypothesis Testing (Single Population)

Autumn 2019

Lecture Week 8: Hypothesis Testing
26134 Business Statistics

© 2019 Paul Burke University of Technology Sydney

The Overall Picture …
 Last Lecture: Chp 8. Interval Estimation
26134 Business Statistics

 This Lecture: Chp 9. Hypothesis Testing 

 Next Lecture: Chp 10: Comparisons Involving Means

2 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 1

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

We Have A Problem … You’re Too Slow
 You are regional manager of operations for CFK, a small take 
away food franchise
 Head office at CFK have had complaints about how slow drive‐
through operations are at one of your stores so Jenny from 
26134 Business Statistics

head‐office comes out to investigate and finds:
 It took an average of 5.6 minutes from 
customers ordering to receiving their food and 
driving off (with a standard deviation of 2.4 
minutes)
 She based both these statistics on a sample of 
41 observations
 Jenny reminds you that the company has a 
benchmark that customers should be served, 
on average, in under 5 minutes
How are you going to defend yourself?
3 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Executing the Innocent: A Lesson for Jenny
What is Jenny’s claim focused on? On what basis are Jenny’s claims made?

Unknown Population Parameter


26134 Business Statistics

Statistical
Inference

x Sample Statistic

Observable;
from n
observations

L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 2

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Managerial Steps: HTAB System to Test Hypotheses 
eg. CFK drive through operations
HYPOTHESISE Your store Steve’s store
About population parameters Ho: μ ≤ 5min Ho: μ ≤ 5min
Ha: μ > 5min Ha: μ > 5min
26134 Business Statistics

TEST Sample of n=41; Sample of n=41;

Gather sample statistics xbar = 5.6 mins xbar = 6.6 mins

Compare to Compare to
ACTION:  critical value that allows
critical value that allows
Reject or Do Not Reject for sampling error for sampling error
Do not reject Ho: μ≤5 Reject Ho: μ≤5

Do not change Change operations

BUSINESS 
operations e.g. Re-train Staff
IMPLICATIONS
Keep your job! $$
5 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Steps we will undertake today
1. State the (null and alternative) hypotheses

2. Determine the critical value (and associated 
26134 Business Statistics

decision‐rule by which you will reject the null)

3. Determine the test‐statistic (the evidence you 
have gathered from a sample is standardized)

4. Make a statistical conclusion (reject Ho or do not 
reject)  and managerial conclusion
Note: Best that step 2 comes before step 3 to avoid being accused of being biased/unethical.
6 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 3

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

26134 Business Statistics

1. State the (Null & Alternative) Hypotheses

7 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

A specific value and two outcomes 
= Two competing Hypotheses
H : parameter relationship value
Competing H : parameter opposing relationship value
 Jenny wants to know if 
waiting at KFCs drive  H:
through, on average,  Competing H :
significantly exceeds five 
26134 Business Statistics

minutes.

 Jill wants to know if the 
average distance travelled 
by her sales team is  H:
significantly less than  Competing H :
expected distance of 
100km

 Jim wants to know if the 
new product being  H:
produced by a 3D printer 
is exactly 300 grams, on  Competing H :
average.

8 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 4

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Task 1:
HYPOTHESISE
Two Hypotheses
 Null Hypothesis (H0):
 A maintained hypothesis that is held to be true unless
sufficient evidence to the contrary is obtained
26134 Business Statistics

 Alternative hypothesis (Ha):

 A hypothesis against which the null hypothesis is tested and
which will be held to be true if the null is held false.
 We often associate the word “significant” (e.g., something is
significantly different; significantly more than, significantly
less than) with the alternative.
 It usually implies that by rejecting the null, we take action.

Source: Newbold et al. 2003, p.311.

9 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Developing Null & Alternative Hypotheses: Hung 
up on HA!
 “Determining the null and alternative hypothesis is often a 
difficult task … The null hypothesis represents the situation 
assumed to be true unless the evidence is strong enough 
26134 Business Statistics

to convince the decision maker it is not true.” (Groebner et 
al. 2006).
 “… whatever we’re investigating is specified as the 
alternative (research) hypothesis” (Keller & Warrack)
 “The alternative hypothesis is what the test is attempting 
to establish” (Anderson et al.)
 “The alternative hypothesis, denote HA, is really the more 
important one, because it is the hypothesis that answers 
our question.” (Selvanathan et al.)

10 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 5

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

The Proof by Contradiction Philosophy
 Warning: You can often get into difficulty when 
they try and use the following philosophy:
 “You must always form a hypothesis such that you 
26134 Business Statistics

hope to find evidence to reject the null and favour 
the competing alternative hypothesis”

So is there an easier way to determine which

hypothesis should be labelled “Ho: - the null
hypothesis” and that the competing hypothesis should
therefore be labelled the “Ha: - the alternative”?

11 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Task 1:
HYPOTHESISE
Overcoming the Difficulty
 “The null hypothesis, Ho, must specify that the population 
parameter is equal to a single value” (Selvanathan et al.)
26134 Business Statistics

Ho:    Ho:  =  Ho:   

Ha:  <  Ha:    Ha:  > 
One-tailed (lower tail) Two-tailed One-tailed (upper tail)

 Ha: implies something is …
“significantly” less, significantly different or significantly more

NOTE: In these examples, the number “5” is used – it could be any numerical value of
interest to the manager
12 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 6

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Three Different Forms of a Hypothesis Test 
About a Population Mean
26134 Business Statistics

  5  = 5   5

a < 5 a  5 a > 5
One-tailed (lower tail) Two-tailed One-tailed (upper tail)

 The rejection region (described by Ha) is the area that 
gets shaded.

NOTE: In these examples, the number “5” is used – it could be any numerical value of
interest to the manager
13 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

A specific value and two outcomes 
= Two competing Hypotheses
H : parameter relationship value
Competing H : parameter opposing relationship value
 Jenny wants to know if 
waiting at KFCs drive  H:
through, on average,  Competing H :
significantly exceeds five 
26134 Business Statistics

minutes.

 Jill wants to know if the 
average distance travelled 
by her sales team is  H:
significantly less than  Competing H :
expected distance of 
100km

 Jim wants to know if the 
new product being  H:
produced by a 3D printer 
is exactly 300 grams, on  Competing H :
average.

14 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 7

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Task 1:
HYPOTHESISE
An Example
 A mobile phone company is concerned about the 
battery life of its batteries being supplied 
 The company’s expectations is that the life of a 
26134 Business Statistics

battery should, on average, be longer than 1095 
days (3 years)
 So in testing, we want to take action (e.g., check 
material inputs, processes, storage conditions, 
etc.) when batches are significantly less than 
1095 days
 What is one hypothesis being asked about battery life?
 What is a competing hypothesis?
 Which of these two hypotheses is the null?
 Which is the alternative?

15 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney
26134 Business Statistics

2. Determine the Critical Value

16 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 8

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

One‐Tailed Test: Basic Idea
One-tailed (upper tail)

Non-rejection Rejection
region region
26134 Business Statistics

1-α = 0.9500 α = 0.05


CRITICAL
Ho:   mins. VALUE Ha:  > mins
NON-REJECTION REGION REJECTION REGION
If the observed statistic from sample If a computed sample statistic (xbar)
falls in this region the decision is to lies in this portion of the distribution,
fail to reject the null hypothesis the null hypothesis will be rejected

17 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Example 1 of 3: One‐Tailed Test
One-tailed (upper tail)

Non-rejection Rejection
region region
26134 Business Statistics

1-α = 0.9500 α = 0.05

 5.7
CRITICAL
Ho:   mins. VALUE
Ha:  > mins
Hypothesise drive through average to be  5 mins. Allowing for some error in relying on
inferences using a sample statistic, your critical value is determined to be 5.7 minutes.
Sample n=100 observations and observe a sample mean, 𝒙 𝟒. 𝟏 𝐦𝐢𝐧𝐮𝐭𝐞𝐬
What would you do (Do Not Reject Ho or Reject Ho)? What would you conclude about
the population?

18 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 9

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Example 2 of 3: One‐Tailed Test
One-tailed (upper tail)

Non-rejection Rejection
region region
1-α = 0.9500
26134 Business Statistics

α = 0.05

 5.7

Ho:   mins. CRITICAL Ha:  > mins

VALUE
Hypothesise drive through average to be  5 mins. Allowing for some error in relying on
inferences using a sample statistic, your critical value is determined to be:5.7 minutes
Sample n=100 observations and observe a sample mean, 𝒙 𝟓. 𝟓 𝐦𝐢𝐧𝐮𝐭𝐞𝐬
What would you do (Do Not Reject Ho or Reject Ho)? What would you conclude about
the population?

19 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Example 3 of 3: One‐Tailed Test
One-tailed (upper tail)

Non-rejection Rejection
region region
26134 Business Statistics

1-α = 0.9500 α = 0.05

 5.7

Ho:   mins. CRITICAL Ha:  > mins

VALUE
Hypothesise drive through average to be  5 mins. Allowing for some error in relying on
inferences using a sample statistic, your critical value is determined to be:5.7 minutes
Sample n=100 observations and observe a sample mean, 𝒙 𝟓. 𝟕𝟖 𝐦𝐢𝐧𝐮𝐭𝐞𝐬
What would you do (Do Not Reject Ho or Reject Ho)? What would you conclude about
the population?

20 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 10

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Reducing Alpha Requires Stronger Evidence to 
Reject the Null… Rejection Region/Area
One-tailed (upper tail)
=.05
26134 Business Statistics

 5.7
Imagine taking a sample of CRITICAL
Ho:    Ha:  > 
n=100 observations and VALUE
observing a sample mean,
Rejection Region/Area =.01
𝒙 𝟓. 𝟕𝟖 𝐦𝐢𝐧𝐮𝐭𝐞𝐬
What would be the
conclusion when =.05?
When  =.01?

 5.8
21
CRITICAL
L08 – Hypothesis Testing (with Paul Burke)
VALUE© 2019 University of Technology Sydney

How do we determine the Critical Value? 
Rejection
Non-rejection
region
region
α = 0.05
1-α = 0.9500
26134 Business Statistics


Ho:   mins. CRITICAL Ha:  > 
VALUE
SOLUTION:
(1) Assume =.05 unless told otherwise (as per front of exam booklet)
(2) We know area in upper tail ( or /2), we know sample size minus 1
(i.e., df=n-1) – we use intersection of two to look up t-tables and find ‘t’
(3) This value of ‘t’ tells us how many standard errors ( s/sqrt(n) ) to go
above (or below) the hypothesised value (0) before start rejecting Ho.
L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 11

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

26134 Business Statistics Critical Values from the Student-t Distribution

df=(n-1)

One-tail test Two-tail test One-tail test Two-tail test

for =.05 for =.05 for =.01 for =.01
23 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Critical Values from the Student-t Distribution

26134 Business Statistics

df=(n-1)

One-tail test Two-tail test One-tail test Two-tail test

for =.05 for =.05 for =.01 for =.01
24 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 12

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Table of Common Critical Values
Notice last two entries of t-table when n is large … ?

Shows that Z = t when df =n-1 approach infinity

26134 Business Statistics

So we can use these shortcut values if ‘s’ is unknown but n is

large – this will depend on what value of d.f. your t-table has.

% Alpha One-tailed Two-Tailed

% sign.
confidence   
0.99 0.01 2.326 2.576
* 0.95 0.05 1.645 1.96
0.90 0.10 1.282 1.645
0.80 0.20 0.842 1.282

25 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Example of Finding Critical Value
Non-rejection Rejection
region region
1-α = 0.9500 α = 0.05
26134 Business Statistics


Ho:   mins. CRITICAL
Ha:  > 
VALUE
Example, suppose we wish to find test-statistic when a sample size of
41 observations reveals a mean of 5.6 minutes with a standard
deviation of 2.4 minutes in drive through deliveries.
(1) Obtain the critical value: tα or α/2, df=n-1 =

L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 13

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

OBTAINING THE
CRITICAL VALUE
Battery Example
 A mobile phone company wants to test 
hypothesis that a batch of batteries being 
supplied have a battery life that is significantly 
less than 1095 days
26134 Business Statistics

 The company wants to test this hypothesis at the 
99% level (i.e. =.01) and tested the life of 201 
batteries in the batch to do so
 What is the critical value that will be used to test the 
hypothesis?

27 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney
26134 Business Statistics

3. Determine the Test‐Statistic

28 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 14

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

One‐Tailed Test: Basic Idea
One-tailed (upper tail)

Non-rejection Rejection
region region
26134 Business Statistics

1-α = 0.9500 α = 0.05


CRITICAL
Ho:   mins. VALUE Ha:  > mins
NON-REJECTION REGION REJECTION REGION
If the observed statistic from sample If a computed sample statistic (xbar)
falls in this region the decision is to lies in this portion of the distribution,
fail to reject the null hypothesis the null hypothesis will be rejected

29 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Standardizing the Evidence = The Test Statistic
To standardize anything:
(a) subtract the mean of distribution (in this case the hypothesised value)
(b) divide by the standard error of distribution.
26134 Business Statistics

Evidence  Hypothesised value x

Test statistic   t
Standard error of statistic sx n

• It is a unit free measure

• Can be compared to critical values
• Critical values come from standardised Student t-
distribution table or standard normal (z) distribution
tables.
tα or α/2, df=n-1

30 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 15

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Example of Obtaining Test‐Statistic
Non-rejection Rejection
region region
1-α = 0.9500 α = 0.05
26134 Business Statistics


Ho:   mins. CRITICAL
Ha:  > 
VALUE
Example, suppose we wish to find test-statistic when a sample size of
41 observations reveals a mean of 5.6 minutes with a standard
deviation of 2.4 minutes in drive through deliveries.
(2) Obtain the test-statistic:
x
t
sx n
L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

OBTAINING THE
TEST-STATISTIC
Battery Example
 A mobile phone company wants to test 
hypothesis that a batch of batteries being 
supplied have a battery life that is significantly 
less than 1095 days
26134 Business Statistics

 The company wants to test this hypothesis at the 
99% level (i.e. =.01) and tested the life of 201 
batteries in the batch to do so
 The sample had an average life of 1090 days and 
a sample standard deviation of 28.4 days
 What is the test‐statistic that will be used to test the 
hypothesis?

32 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 16

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

26134 Business Statistics

4. Draw a (Statistical & Managerial) Conclusion

33 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Interpretation:
Non-rejection Rejection
region region
1-α = 0.9500 α = 0.05
26134 Business Statistics


Ho:   mins. CRITICAL
Ha:  > 
VALUE
Example, suppose we wish to find test-statistic when a sample size of
41 observations reveals a mean of 5.6 minutes with a standard
deviation of 2.4 minutes in drive through deliveries.
• Are you going to the reject the null hypothesis?
• If so, what are the business implications?
• Can you keep your job (at least for the moment)?

L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 17

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

MAKING A
Battery Example CONCLUSION
 1. A mobile phone company wants to test 
hypothesis that a batch of batteries being 
supplied have a battery life that is significantly 
less than 1095 days. As such, you hypothesised: 
26134 Business Statistics

 Ho: µ1095 days; 
 Ha: µ < 1095 days
 The company wants to test this hypothesis at the 
99% level (i.e. =.01) and tested the life of 201 
batteries in the batch to do so.
 2. You obtained critical value of:  __________
 The sample had an average life of 1090 days and 
a sample standard deviation of 28.4 days
 3. You obtained a test‐statistic of: _________
 4a. Will you reject the null hypothesis?
 4b. What is the managerial conclusion you can reach 
about the adequacy of the batteries?
35 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Summary of Steps
 1a. Determine the hypothesis
 One‐tailed if null hypothesis ≤ or ≥; two tailed if null has =
 1b. Sketch and shade rejection region indicated by Ha
 2a. Specify the level of significance
 If not given, assume =.05
26134 Business Statistics

 2b. Select the distribution and obtain critical value from tables
 If standard deviation is known, use z
 If standard deviation is unknown, use a t with d.f.=n‐1
 3. Examine the evidence
 Can standardize evidence (called a test‐statistic) putting it into 
standardized units where t=(sample mean – hyp value)/std.error
 Can convert to original units via: hyp value +/‐(table)*(margin of error)
 4. Determine whether to Reject Ho:
 Reject if evidence falls in rejection region
 Shortcut is abs(test‐statistic) > critical value, but double check 
relationship of sample mean to hypothesised mean
* As long as you compare apples to apples and not apples to oranges, 
36 L08 – Hypothesis Testing (with Paul Burke)
both methods will reach the same conclusion!
© 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 18

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Application: A Problematic 3‐D Printer
 Jochen is an entrepreneur at UTS who has developed 
a new toy for children
 Jochen wants to know if the new product being 
produced by a 3D printer is not significantly different 
26134 Business Statistics

from a target weight of 300 grams, on average, as he 
has simulated in his designs and achieved using 
another printer in the past.
 Jochen takes a sample of 41 new prototypes to be 
weighed by a specialist machine
 The sample prototypes weigh 294 grams on average
with a standard deviation of 4 grams
 What should Jochen conclude?
37 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

3‐D Printer Problem
 Hypotheses?

Critical value?
26134 Business Statistics

 Evidence/test‐statistic?

 Comparison? Reject Ho?

 Conclusion?

38 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 19

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

26134 Business Statistics

A Summary of Decision Rules …

39 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Summary: One‐Tailed Test (Upper)
Total Rejection Region/Area is 
Do not Rej Ho Area
above
=
Rej Ho
26134 Business Statistics

 x
Critical value
Ho:  ≤ o Ha:  > o

Reject Ho if: t  x   > tα, df=n-1 and x   0

sx n
Test-statistic should be positive, so absolute value unnecessary, but useful
to see how shortcut for two-tailed comes about
40 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 20

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Summary: One‐Tailed Tests (Lower)
Total Rejection Region/Area is 
Area below
=

Rej Ho Do not Rej Ho

26134 Business Statistics

 x
- Critical value

Ha:  < o Ho:  ≥ o

Reject Ho if: t  x   > t α , df=n-1 and x   0

sx n

41 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Summary: Two‐Tailed Test Total Rejection

Region/Area is 

Area
below Do not Rej Ho Area
= /2 above
= /2
Rej Ho Rej Ho
26134 Business Statistics

 x
- Critical value Critical value

Ha:  < o Ho:  = o Ha:  > o

Reject Ho if: t  x   > tα/2, df=n-1

sx n

42 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 21

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Standardising the Comparison
 We could compare the critical value (in dollars) to 
the evidence (sample mean in dollars) 
26134 Business Statistics

 If the sample mean lies in the rejection region, we 
reject the null
 If the sample mean lies in the non‐rejection region, 
we do not reject the null
 In practice, we instead compare 
 the critical value (from the student‐t tables) to the 
sample mean that has been standardized … this is 
called our test‐statistic

43 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Elements of a Hypothesis Test
We can also test by comparing
critical value to evidence (dollars
to dollars; minutes to minutes):
26134 Business Statistics

Hypothesised Critical Evidence Dollars or

value value
|
Mins.
| |
0 Z or t Test
(1) table statistic Standard
(2) (3) Errors or t/Z-
scores

44 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 22

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

Example of Critical Value in Original Units:
Non-rejection Rejection
region region
1-α = 0.9500 α = 0.05
26134 Business Statistics


Ho:   mins. CRITICAL
Ha:  > 
VALUE
Example, suppose we wish to find test-statistic when a sample size of
41 observations reveals a mean of 5.6 minutes with a standard
deviation of 2.4 minutes in drive through deliveries.
(3b) As a check on answer, you can obtain the critical value: tα or
α/2, df=n-1 and convert back to original units:
 s   2.4 
   t or  / 2 , n 1 * x   5  1 .684 *    5 .631 mins
 n  41 
L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Next Week
 We continue on with hypothesis testing
 Strongly recommend you revise over the next 
26134 Business Statistics

week on this important topic!

 Next week:
 Hypothesis to compare two means:
 Paired sample t‐test 
 Independent samples t‐test

 Hypothesis to compare two or more means:
 ANOVA (stands for Analysis of Variance)

46 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 23

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

26134 Business Statistics

Tests Involving Population Mean Proportions
NOTE: This is topic is for noting only.
This will not be assessable (i.e., you won’t see such a
question in the exam or the assignment).
It is here to complete your notes for future applications
should you so need.
47 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Hypothesis Test: Population Proportion
The process is the SAME.
All that differs is the standard error.
26134 Business Statistics

Hypothesised Critical Test Hypothesised Critical Test

value value statistic value value statistic
Ho:  ≤ o; Ha:  > o Ho: p  po ; Ha: p > po
x
t z=
sx n
48 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 24

26134 Business Statistics Hypothesis Testing (Single Population)
Autumn 2019

One / Two‐Tailed Test: Population Proportion
 Assume np and nq >= 5 in order to use z
 Still three potential hypotheses
 Ho: p  po ; Ha: p > po
26134 Business Statistics

 Ho: p  po ; Ha: p < po
 Ho: p = po ; Ha: p  po
 Still specify level of significance (e.g., =.05)
 Only noticeable difference is the standard error of 
the hypothesised mean proportion (therefore 
known):
p o (1  p o ) pˆ qˆ
 pˆ  
n n
49 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Hypothesis Testing Involving Mean Proportions
po (1  po ) … is the standard error and p0 is the
 pˆ  hypothesised mean proportion
n
Standardized evidence from data = Test statistic becomes:
26134 Business Statistics

z=

Critical value = zα or α/2 from the z-tables (we have enough

information about standard deviation from sample mean)

z=
Reject Ho if: > Zα or α/2
And that p p has sign as intended.

50 L08 – Hypothesis Testing (with Paul Burke) © 2019 University of Technology Sydney

Dr. Paul F. Burke @drpaulburke 25