You are on page 1of 373

METHODS AND EXCEL-BASED APPLICATIONS 4 t h E d it io n

Applied Business Statistics


Applied Business
Statistics
METHODS AND EXCEL-BASED APPLICATIONS
4th Edition

TREVOR WEGNER

SOLUTIONS MANUAL

www.jutaacademic.co.za
TREVOR WEGNER
Applied Business Statistics: Methods and Excel-based Applications: Solutions Manual

Print edition first published in 1993


Reprinted 2000 and 2003
Second Edition 2008
Third Edition 2012
Fourth edition 2015 (Web PDF)

Juta and Company (Pty) Ltd


First Floor
Sunclare Building
21 Dreyer Street
Claremont
7708

PO Box 14373, Lansdowne 7779, Cape Town, South Africa

© 2015 Juta & Company (Pty) Ltd

ISBN 978 1 48511 788 9 (Web PDF)

All rights reserved. No part of this publication may be reproduced or transmitted in any form or
by any means, electronic or mechanical, including photocopying, recording, or any information
storage or retrieval system, without prior permission in writing from the publisher. Subject to
any applicable licensing terms and conditions in the case of electronically supplied
publications, a person may engage in fair dealing with a copy of this publication for his or her
personal or private use, or his or her research or private study. See section 12(1)(a) of the
Copyright Act 98 of 1978.

The author and the publisher believe on the strength of due diligence exercised that this work
does not contain any material that is the subject of copyright held by another person. In the
alternative, they believe that any protected pre-existing material that may be comprised in it
has been used with appropriate authority or has been used in circumstances that make such
use permissible under the law.
CHAPTER 1

STATISTICS IN MANAGEMENT

1.1 It is a decision support tool. It generates evidence based information through analysis of
data to inform management decision making.

1.2 Descriptive statistics summarises (profiles) sample data; inferential statistics generalises
sample findings to a broader population (to estimate values or confirm relationships).

1.3 Statistical modelling is explores and quantifies relationships between variables for
estimation or prediction purposes.

1.4 Data quality is influenced by: (i) Data source; (ii) Data collection method; and (iii) Data type

1.5 Different statistical methods are valid for different data types.

1.6 In data preparation, consider (i) Data relevancy; (ii) Data cleaning; and (iii) Data enrichment.

1.7 (a) Random variable: Performance appraisal system used


(b) Population: All JSE companies
(c) Sample: The 68 HR managers surveyed
(d) Sampling unit: a JSE-listed company
(e) 46% is a sample statistic
(f) Random sampling is necessary to allow valid inferences to be drawn based on the sample
evidence.

1.8 (a) Random variable: Female magazine readership


(b) Population: All female magazine readers
(c) Sample: The 2000 randomly selected female magazine readers
(d) Sampling unit: a female reader of a female magazine
(e) 35% (700/2000) is a sample statistic
(f) Inferential statistics – as its purpose is to test the belief that market share = 38%

1.9 (a) Three (3) random variables. They are:


(i) weekly sales volume; (ii) number of ads placed per week; (iii) advertising media used.
(b) Dependent variable = weekly sales volume
(c) Independent variables = number of ads placed per week; advertising media used.
(d) Statistical model building (predict sales volume from ads placed and media used)
1.10 Scenario 1 Inferential statistics
Scenario 2 Descriptive statistics
Scenario 3 Descriptive statistics
Scenario 4 Inferential statistics
Scenario 5 Inferential statistics
Scenario 6 Inferential statistics
Scenario 7 Inferential statistics

1.11 (a) numeric, ratio-scaled, continuous {21,4 years; 34,6 years}


(b) numeric, ratio-scaled, continuous {416,2m²; 3406,8m²}
(c) categorical, ordinal-scaled, discrete {matric; diploma}
(d) categorical, nominal-scaled, discrete {married; single}
(e) categorical, nominal-scaled, discrete {Boeing; Airbus}
(f) categorical, nominal-scaled, discrete {verbal; emotional}
(g) numeric, ratio-scaled, discrete {41 ; 62}
(h) categorical, ordinal-scaled, discrete {salary only; commission only}
(i) (i) categorical, ordinal-scaled, discrete {1 = apple; 2 = orange}
(ii) categorical, nominal-scaled, discrete {yes; no}
(iii) categorical, nominal-scaled, discrete {train; bus}
(iv) numeric, interval-scaled, discrete {2; 5}
(j) numeric, ratio-scaled, continuous {12,4kg; 7,234kg}
(k) categorical, nominal-scaled, discrete {Nescafe; Jacobs}
(l) numeric, ratio-scaled, continuous {26,4 min; 38,66 min}
(m) categorical, ordinal-scaled, discrete {Super; Standard}
(n) numeric, ratio-scaled, continuous {R85,47; R2315,22}
(o) numeric, ratio-scaled, discrete {75; 23}
(p) numeric, ratio-scaled, discrete {5; 38}
(q) numeric, ratio-scaled, continuous {9,54 hours; 10,12 hours}
(r) numeric, interval-scaled, discrete {2; 6}
(s) numeric, ratio-scaled, discrete {75; 238}
(t) categorical, nominal-scaled, discrete {Growth funds; Industrial funds}

1.12 (a) 11 random variables


(c) Illustration value
(b) Economic sector categorical, nominal-scaled, discrete {retail}
Head office region categorical, nominal-scaled, discrete {Gauteng}
Company size numeric, ratio-scaled, discrete {242}
Turnover numeric, ratio-scaled, continuous {R3 432 562}
Share price numeric, ratio-scaled, continuous {R18.48}
Earnings per share numeric, ratio-scaled, continuous {R2.16 / share}
Dividends per share numeric, ratio-scaled, continuous {R0.86 / share}
Number of shares numeric, ratio-scaled, discrete {12 045 622}
ROI (%) numeric, ratio-scaled, continuous {8.64%}
Inflation index (%) numeric, ratio-scaled, continuous {6.75%}
Year established numeric, ratio-scaled, discrete {1988}
1.13 (a) 13 random variables
(c) Illustration value
(b) Gender categorical, nominal-scaled, discrete {female}
Home language categorical, nominal-scaled, discrete {Xhosa}
Position categorical, ordinal-scaled, discrete {middle manager}
Join date numeric, ratio-scaled, discrete {1998}
Status categorical, ordinal-scaled, discrete {gold status}
Claimed categorical, nominal-scaled, discrete {yes}
Problems categorical, nominal-scaled, discrete {yes}
Yes problem categorical, nominal-scaled, discrete {online access difficult}
Services - airlines numeric, interval-scaled, discrete {2}
Services – car rentals numeric, interval-scaled, discrete {5}
Services - hotels numeric, interval-scaled, discrete {4}
Services – financial numeric, interval-scaled, discrete {2}
Services – telecomms numeric, interval-scaled, discrete {2}

1.14 Financial Analysis data: mainly numeric (quantitative), ratio-scaled.


Voyager Services Quality data: mainly categorical (qualitative); but when rating scales
are used, such as in Question 8, the data is numeric, but interval-scaled and discrete.

---ooOoo---
CHAPTER 2

EXPLORATORY DATA ANALYSIS

SUMMARISING DATA SUMMARY TABLES AND GRAPHS

Exercise 2.1 A picture is worth a thousand words.

Exercise 2.2 (a) bar (or pie) chart


(b) multiple (or stacked) bar chart
(c) histogram
(d) scatter plot

Exercise 2.3 Cross-tabulation table (or joint frequency table; or two-way pivot table).

Exercise 2.4 Bar chart (i) displays data on a categorical variable


(ii) categories can be displayed in any order
(iii) width of bars is arbitrary (but all of equal widths)

Histogram (i) displays numerical data only


(ii) intervals must be continuous (and constant width) and sequential
(iii) width of bars is determined by interval width

Exercise 2.5 Line graph


Exercise 2.6 File: X2.6 - magazines.xlsx

(a) Magazine preferences by female teenagers

Magazine Count %
True Love 95 19%
Seventeen 146 29%
Heat 118 24%
Drum 55 11%
You 86 17%
Total 500 100%

Percent of Female Teenagers

17% 19% True Love


Seventeen
11%
Heat
Drum
29%
24% You

(b) Interpretation

Seventeen is the most popular teenager magazine (29% of female teenager prefer it).
Almost a quarter of the female readers surveyed prefer reading Heat (24%), while the least
preferred magazine is Drum with only 11% of female magazine readers preferring it.
Exercise 2.7 File: X2.6 - magazines.xlsx

(a) Magazine preferences by female teenagers

Magazine Count %
True Love 95 19%
Seventeen 146 29%
Heat 118 24%
Drum 55 11%
You 86 17%
Total 500 100%

% of Female Teenagers per Magazine

35%
29%
30%
24%
25%
19%
20% 17%
15% 11%
10%
5%
0%

(b) Heat is preferred by 24% of all female teenager readers.


Exercise 2.8 File: X2.8 - job grades.xlsx

(a) and (b) Categorical Frequency Table - Job Grades

Job grade Data Total


A Count 14
% 35% Job Grade %
B Count 11 A 35
% 27.5% B 27.5
C Count 6 C 15
% 15% D 22.5
D Count 9 Total 100
% 22.5%
Total Count 40
Total % 100%

(c) 22.5% of employees are in job grade D

(d) Bar Chart and Pie Chart - Job Grades

% Employees per Job Grade


% of Employees per Job Grade

40
35
35
30 27.5

25 22.5
D A A
20 23% 35%
15 B
15 C C
15% B
10 D
27%
5
0
A B C D
Exercise 2.9 File: X2.9 - office rentals.xlsx

(a) and (b) Numerical Frequency Distribution and Cumulative Frequency Distribution

Rentals Count % Count Cum %


≤ 200 4 13.3 13.3%
201 - ≤250 8 26.7 40.0%
251 - ≤300 9 30 70.0%
301 - ≤350 6 20 90.0%
351 - ≤400 3 10 100.0%
More 0 0 100.0%
Total 30 100

(c) (i) 13.3% of all office space costs less than or equal to R200 / m2
(ii) 70% of all office space costs at most R300 / m²
(iii) 10% of all office space costs more than R350 / m²
(iv) 9 office buildings have rentals between R300 and R400 / m²
Exercise 2.10 File: X2.10 - storage dams.xlsx

Cape Town water storage dams capacities

(a) Storage Dam Capacity (Ml) %


Wemmershoek 158644 16.9
Steenbras 95284 10.2
Voelvlei 244122 26
Theewaterskloof 440255 46.9
Total capacity 938305 100

Capacity of Cape Town Storage Dams


(in Million litre)

17% Wemmershoek
47% 10% Steenbras
Voelvlei
26% Theewaterskloof

(b) (i) Voelvlei dam supplies 26% of Cape Town's water.


(ii) Wemmershoek and Steenbras dams together provide 27.1% of Cape Town's water.
Exercise 2.11 File: X2.11 - taste test.xlsx

(a) Taste test preferences for fruit juices

Blind Label Brand Number %


A Liqui Fruit 45 18.0
B Fruiti Drink 26 10.4
C Yum Yum 64 25.6
D Fruit Quencher 38 15.2
E Go Fruit 77 30.8
Total 250 100.0

Bar Chart - Fruit Juice Preferences

Consumer Preferences for Fruit Juice Brands

40.0

35.0
30.8
30.0
25.6
% of consumers

25.0

20.0 18.0
15.2
15.0
10.4
10.0

5.0

0.0
Liqui Fruit Fruiti Drink Yum Yum Fruit Quencher Go Fruit

Pie Chart - Fruit Juice Preferences

Consumer Preferences - Fruit Juices

45, 18%
Liqui Fruit
77, 31%
Fruiti Drink
26, 10% Yum Yum
Fruit Quencher

38, 15% Go Fruit


64, 26%

(b) 18% of the sampled consumers prefer Liqui Fruit.

(c) 56.4% of the sampled consumers prefer either Yum Yum or Go Fruit.
Exercise 2.12 File: X2.12 - annual car sales.xlsx

Manufacturer Annual Sales % Sales


Toyota 96959 19.4
Nissan 63172 12.6
Volkswagen 88028 17.6
Delta 62796 12.6
Ford 74155 14.8
MBSA 37268 7.5
BMW 51724 10.4
MMI 25354 5.1
Total Sales 499456 100.0

(a) Bar Chart - Annual Car Sales by Manufacturer

Annual Sales of Passenger Cars by Manufacturer

120000

96959
100000
88028

80000 74155
63172 62796
60000 51724

37268
40000
25354
20000

(b) Percentage Pie Chart - Annual Car Sales by Manufacturer

% Annual Sales of Passenger Cars by Manufacturer


MMI, 5.1

BMW, Toyota
10.4 Toyota, 19.4 Nissan
MBSA, 7.5
Volkswagen
Delta
Nissan, 12.6
Ford, 14.8 Ford
MBSA
Volkswagen,
BMW
Delta, 12.6 17.6
MMI

(c) Total % held by top three manufacturers - Toyota (19.4%), Volkswagen (17.6%)
and Ford (14.8%) - represents 51.8% of the total passenger car market.
Exercise 2.13 File: X2.13 - half-yearly car sales.xlsx

Manufacturer First half Second half % Change


Toyota 42661 54298 27.3
Nissan 35376 27796 -21.4
VW 45774 42254 -7.7
Delta 26751 36045 34.7
Ford 32628 41527 27.3
MBSA 19975 17293 -13.4
BMW 24206 27518 13.7
MMI 14307 11047 -22.8

(a) Multiple bar chart - Car Sales by Half-Year and Manufacturer

Half-yearly Car Sales by Manufacturer

60000

50000

40000
First half
30000
Second half
20000

10000

0
Toyota Nissan VW Delta Ford MBSA BMW MMI
First half 42661 35376 45774 26751 32628 19975 24206 14307
Second half 54298 27796 42254 36045 41527 17293 27518 11047

(b) First half-year best performers: Nissan; Volkswagen; MBSA and MMI

(c) Delta showed the largest % increase from the first half to the second half of 34.7%.
Refer to the above Table for the % Change between First and Second Half-Year Sales.
Exercise 2.14 File: X2.14 - television brands.xlsx

(a) Categorical Frequency Table - Television Brands Owned

Count of Brands
Brands Total
Daewoo 16%
LG 30.4%
Philips 10.4%
Sansui 24%
Sony 19.2%
Grand Total 100%

(b) Percentage Bar Chart - Television Brands Owned

% of TV Brands Owned

35%
30.4%
30%
Brands Total 24%
25% Daewood 16%
20% LG 30.4% 19.2%
Philips
16% 10.4%
15% Sansui 24%
10.4%
10%
Sony 19.2%
Total 100%
5%

0%
Daewood LG Philips Sansui Sony

(c) Philips is the least preferred brand (preferred by only 10.4% of households surveyed).

(d) The most popular brand is LG that is owned by 30.4% of the households surveyed.
Exercise 2.15 File: X2.15 - estate agents.xlsx

(a) Frequency Count Table


Count of House sales
House sales Total
3 12
4 15
5 6
6 7
7 5
8 3
Grand Total 48

(b) Histogram - Residential Properties Sold per Estate Agent

Histogram of Residential Properties Sold per Agent


16

14

12

10

8
15
6 12
4
6 7
2 5
3
0
3 4 5 6 7 8
Total 12 15 6 7 5 3

(c) The most frequently sold number of properties per estate agent was 4.
4 properties each were sold by (15/48) 31.3% of all estate agents

(d) The same frequency count table (a) and histogram (b) is produced.
Exercise 2.16 File: X2.16 - fast foods.xlsx

Fast Food Outlet Count % Fast Food %


KFC 56 17.2 KFC 17.2
St Elmo's 58 17.8 St Elmo's 17.8
Steers 45 13.8 Steers 13.8
Nandos 64 19.7 Nandos 19.7
Ocean B 24 7.4 Ocean B 7.4
Butler's 78 24.0 Butler's 24
Total 325 100

(a) Percentage Bar Chart - Consumer Preferences of Fast Food Outlet

Percentage of Consumers
30.0
24
25.0
19.7
20.0 17.2 17.8
13.8
15.0
%
10.0 7.4
5.0
0.0
KFC St Elmo's Steers Nandos Ocean B Butler's

(b) Percentage Pie Chart - Consumer Preferences of Food Type

Firstly produce the categorical frequency table of Food Type Preferences. Sum
the frequency counts of the different food types (e.g. Chicken = 56 + 64 = 120) and
express the total count as a % of the total number of customers (e.g. 120/325 = 36.9%).

Categorical Frequency Table - Consumer Preferences of Food Type

Food type %
Chicken 36.9
Pizza 41.8
Burger 13.8
Fish 7.4
Consumer preference (%) by Food Type
Fish, 7.4

Burger, 13.8
Chicken
Chicken,
36.9 Pizza
Burger
Pizza, 41.8
Fish

(c) Brief Report


Pizza (42%) and Chicken (37%) dominate almost 80% of
the fast food market with Pizzas being slightly more favoured by fast food consumers.
Exercise 2.17 File: X2.17 - airlines.xlsx

(a) Two-way Pivot Table - Counts, Row % (by Airline), Column % (by Passenger)

Passenger
Airline Data Business Tourist Grand Total
Comair Count of Passenger 12 8 20
Row % 60.0% 40.0% 100%
Column % 33.3% 23.5% 28.6%
Total % 17.1% 11.4% 28.6%
Kulula Count of Passenger 4 16 20
Row % 20.0% 80.0% 100.0%
Column % 11.1% 47.1% 28.6%
Total % 5.7% 22.9% 28.6%
SAA Count of Passenger 20 10 30
Row % 66.7% 33.3% 100.0%
Column % 55.6% 29.4% 42.9%
Total % 28.6% 14.3% 42.9%
Total Count of Passenger 36 34 70
Total Row % 51.4% 48.6% 100%
Total Column % 100% 100% 100%
Total Total % 51.4% 48.6% 100%

(b) Two-way Pivot table - Row Percentage by Airline

Count of Passenger Passenger


Airline Business Tourist Grand Total
Comair 60.0% 40.0% 100.0%
Kulula 20.0% 80.0% 100.0%
SAA 66.7% 33.3% 100.0%
Grand Total 51.4% 48.6% 100.0%

(c) Multiple Bar Chart - Passenger Type by Airline

Multiple Bar Chart - Airline by Passenger Type


90%
0.8
80%
% of passenger per airline

70% 0.67
0.6
60%
50%
0.4
40% 0.33 Business
30% Tourist
0.2
20%
10%
0%
Comair Kulula SAA
Business 0.6 0.2 0.666666667
Tourist 0.4 0.8 0.333333333

(d) 42.9% of passengers prefer to fly with SAA.


(e) Kulula is most preferred by tourists (47.1% of tourists prefer Kulula).

(f) Not true. Most business travellers prefer SAA (55.6%).


More than half (55.6%) of all business travellers prefer SAA.
Exercise 2.18 File: X2.18 - car occupants.xlsx

(a) Random Variable - Number of occupants per car


Data Type - Numerical, discrete, ratio-scaled

(b) (i) Numeric % Frequency Distribution (and Cumulative Frequencies)

Occupants Count % Cum Count Cum %


1 23 38.3 23 38.3
2 15 25.0 38 63.3
3 10 16.7 48 80.0
4 5 8.3 53 88.3
5 7 11.7 60 100.0
Total 60 100.0

(b) (ii) Histogram - Occupants per Car

Histogram of Occupants per Car

30

25 23

20
No. of Cars

15
15
10
10 7
5
5

0
1 2 3 4 5

Occupants

(b) (iii) "Less-Than" Ogive (see (a) above) and Cumulative Frequency Polygon

Cumulative Frequency Polygon - Car Occupants

70

60 60
53
50 48
No. of Cars

40 38
30
23
20

10

0
1 2 3 4 5

No. of occupants
(c) (i) 38.3% of motorists travel alone.
(ii) 36.7% of vehicles have 3 or more occupants
(iii) 63.3% of vehicles have no more than 2 occupants.
Exercise 2.19 File: X2.19 - courier trips.xlsx

(a) Random variable - distance travelled (in kms) per courier trip
Data type - numerical, continuous, ratio-scaled

(b) (i) (ii) Numeric % Frequency Distribution (and Cumulative Frequencies)

Distance Count % Cum Count Cum %


≤10 4 8 4 8
11 - ≤15 7 14 11 22
16 - ≤20 15 30 26 52
21 - ≤25 12 24 38 76
26 - ≤30 9 18 47 94
31 - ≤35 3 6 50 100
Total 50 100

(b) (iii) Histogram - Courier Travelling Distances per Trip

Histogram of Courier Travelling Distances

18
16 15

14
12
12
No. of trips

10 9

8 7

6
4
4 3

2
0
≤10 11 - ≤15 16 - ≤20 21 - ≤25 26 - ≤30 31 - ≤35

Distance (in km)

Cumulative % Frequency Polygon


Distances Travelled per Trip

Distance Cum %
100 100
10 8 94
80 15 22
76
Percent of trips

20 52
60 25 76
52
30 94
40
35 100
20 22
8
0
10 15 20 25 30 35

Distance (in km)


(c) (i) 18% of deliveries (9 trips) were between 25 km and 30 km.
(ii) 76% of deliveries (38 trips) were within a 25 km radius.
(iii) 48% of deliveries (24 trips) were beyond a 20 km radius.

(iv) Reading off the Cumulative % Frequency Table (or Polygon) above
52% of trips were no more than 20 km from the depot.
(v) The longest 24% of trips were above 25 km.

(d) The percentage of trips above 30 km is only 6%.


Hence there is adherance to the company policy.
Exercise 2.20 File: X2.20 - fuel bills.xlsx

(a) Random variable - monthly fuel expenditure (in Rands)


Data type - numerical, continuous, ratio-scaled

(b) (i) (ii) Numeric % Frequency Distribution (and (d) the Ogive)

Fuel bill Count % Count Cum Count Cum %


≤300 7 14 7 14
301 - 400 15 30 22 44
401 - 500 13 26 35 70
501 - 600 7 14 42 84
601 - 700 5 10 47 94
701 - 800 2 4 49 98
800+ 1 2 50 100
Total 50 100

(b) (iii) Histogram of Fuel costs / motorist (Rand)

Histogram of Motorists' Monthly Fuel Costs

16 15
14 13
12
no. of motorists

10
8 7 7
6 5
4
2
2 1
0
≤300 301 - 400 401 - 500 501 - 600 601 - 700 701 - 800 800+
Fuel bill (in Rands)

(c) 14% (7 motorists) spend between R500 and R600 per month on fuel.

(d) Cumulative % Frequency Polygon - Motorist fuel bills per month.


(See the Ogive in (b) for Cumulative Frequencies)
Cumulative % Frequency Polygon - Fuel Bills per Month
110
100 98 100
90 94
80 84
70 70
60
50
40 44
30
20
10 14
0
≤300 301 - 400 401 - 500 501 - 600 601 - 700 701 - 800 800+

(e) From (d), approx. 77% of motorists spend less than R550 per month on fuel.

(f) From (a) (and (d)), 30% (15 motorists) spend more than R500 per month on fuel.
Exercise 2.21 File: X2.21 - car sales.xlsx

Data Quarters Corsa Sales


1 37
2 25
3 41
4 29
5 31
6 28
7 30
8 36
9 38
10 62
11 53
12 63
13 43
14 39
15 52
16 61
17 58
18 65
19 73
20 52
21 61
22 46
23 49
24 54

(a) Time Line graph - Quarterly Vehicle Sales

Line Graph of Opel Corsa Quarterly Sales

80
70
60
50
units sold

40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

Quarters

(b) Yes, Corsa sales are showing a general upward trend.


Exercise 2.22 File: X2.22 - market shares.xlsx

Data Year VW Toyota


1 13.4 9.9
2 11.6 9.6
3 9.8 11.2
4 14.4 12.0
5 17.4 11.6
6 18.8 13.1
7 21.3 11.7
8 19.4 14.2
9 19.6 16.0
10 19.2 16.9

(a) Trend Line graphs of Market Shares (%) per car type (VW, Toyota)

Market Share (%) Line Graphs

25

20
% market share

15

10

0
1 2 3 4 5 6 7 8 9 10

year

Legend: Top graph - VW; Bottom graph - Toyota

(b) VW shows a higher sales level but at a declining growth rate.


Toyota shows a lower sales level, but at a rising growth rate.

(c) Choice of franchise is not clearcut, but suggest choosing Toyota because
of its more consistent (steady) growth rate.
Exercise 2.23 File: X2.23 - defects.xlsx

(a) Scatter graph - Inspection time (x) vs Defects found (y)

Scatter Graph of Defects against Inspection time

20
18
no. of defects found per batch

16
14
12
10
8
6
4
2
0
20 30 40 50 60 70

inspection time (minutes)

(b) Yes, there appears to be a moderate to strong positive linear relationship


between the inspection time of a batch and the no. of defects found per batch.
Data Consignment Time Defects
AA 48 17
AB 50 9
AC 43 12
AD 36 7
AE 45 8
AF 49 10
AG 55 14
AH 63 18
AI 55 19
AJ 36 6
AK 40 8
AL 46 14
AM 32 10
AN 50 15
AO 42 14
AP 36 8
AQ 48 12
AR 38 8
AS 45 10
AT 30 6
AU 34 9
AV 43 11
AW 53 16
AX 48 16
AY 56 15
AZ 40 12
BA 33 7
BB 35 10
BC 50 16
BD 48 18
Exercise 2.24 File: X2.24 - leverage.xlsx

(a) Scatter Graph - Profit Growth (y) vs Leverage Ratio (x)

Scatter Graph
Profit Growth (y) and Leverage Ratio (x)

160
140
120
profit growth

100
80
60
40
20
0
30 32 34 36 38 40 42 44 46

leverage ratio

(b) Yes, there is a clear moderate to strong positive linear relationship between
the leverage ratio of a company and its profit growth.

Data Leverage Profit Growth


40.8 111
42.3 116
43.2 132
37.9 105
36.2 69
35.6 40
36.4 58
39.5 118
42.6 104
42.1 125
37.8 97
34.4 76
36.5 98
38.3 100
39.3 75
36.4 88
33.5 20
32.4 25
35.4 78
35.4 65
35.7 84
35.2 88
35.3 86
44.9 115
35.9 50
38.0 92
36.7 110
39.2 72
41.1 128
38.7 86
Exercise 2.25 File: X2.25 - roi%.xlsx

(a) Sector Average Std dev


Mining 9.87 4.58
Services 11.33 2.99
Grand Total 10.70 3.78

(b) Service companies have a higher average ROI% (11.33%) than mining companies (9.87%).
The volatility of ROI% amongst service companies (2.99%) is far lower than amongst mining companies (4.58%)
By inspection, there is a high overlap of ROI% between the two sectors (based on a two-standard deviation
interval around each sample mean). Thus it is likely that there is no statistically significant difference in
mean ROI% between the two sectors.

Exercise 2.26 File: X2.26 - product location.xlsx

(a) Shelf position


Aisle Middle Top Total
Front Average 6.08 5.08 5.58
Std dev 0.890 0.622 0.895
Middle Average 4.24 2.74 3.49
Std dev 1.387 0.297 1.232
Back Average 4.66 4.1 4.38
Std dev 1.193 0.648 0.952
Average 4.99 3.97 4.48
Std dev 1.359 1.114 1.327

Based on shelf position alone, middle shelf positions generate higher average sales (R4.99) than top shelf
(b)
positions (R3.97).
Based on aisle location alone, a front-of-store aisle location generates the highest average sales (R5.58), followed
by a back-of-store aisle location (R4.38).
The lowest average sales occur when the product is displayed in a middle-of-store aisle location (R3.49).
In combination, a front-of-store aisle location on a middle shelf position generates the highest average sales
(R6.08), while a top shelf position in a middle-of-store aisle is the least desirable product location with an average
sales volume of only R2.74.
Sales variability is relatively consistent across aisle locations (0.895 to 1.232) as well as between shelf positions
(1.114 to 1.359).
In combination however, sales volumes show highest variability when the product is positioned in a middle shelf
position in a middle-of-store location (1.387) while the lowest variability in sales volumes occur when positioned in
a top shelf position in a middle-of-store aisle location (0.297).
The large differences in average sales volumes (ranging from R6.08 to R2.74) is evidence of a likley statistically
significant difference in sales volumes due to choice of aisle location and shelf positioning.

Recommendation: A middle shelf position in a front-of-store ailse location is the most preferred product
(c)
display location.
Exercise 2.27 File: X2.27 - property portfolio.xlsx

(a) Numeric Frequency Distribution and Cumulative % of NP%

Intervals Count Cumulative %


-5 0 0.0%
-2.5 4 1.2%
0 7 3.4%
2.5 4 4.6%
5 22 11.4%
7.5 87 38.3%
10 109 71.9%
12.5 46 86.1%
15 31 95.7%
17.5 12 99.4%
20 1 100%
More 1 100%
324 100%

Histogram - Net Profit %


120 109 120.0%

100 87 100.0%

80 80.0%
Frequency

60 60.0%
46
40 31 40.0%
22
20 12 20.0%
4 7 4
0 1 1
0 0.0%
-5 -2.5 0 2.5 5 7.5 10 12.5 15 17.5 20 More
Bin NP%

(b) and (c) Type of business usage


Region Commercial Industrial Retail Total
A Average 7.5 4.9 10.2 7.9
Std dev 2.5 3.1 3.6 3.5
Minimum -2.4 -4.2 0.8 -4.2
Maximum 14.6 8.1 18.4 18.4
Count 104 40 70 214
B Average 12.3 6.8 8.5 9.8
Std dev 3.1 4.2 1.8 3.5
Minimum 2.7 -3.4 4.4 -3.4
Maximum 20.3 10.2 13.2 20.3
Count 46 16 48 110
Average 9.0 5.4 9.5 8.6
Std dev 3.5 3.5 3.1 3.6
Minimum -2.4 -4.2 0.8 -4.2
Maximum 20.3 10.2 18.4 20.3
Count 150 56 118 324

(d) Profile of property portfolio:


The company has almost twice as many properties in region A (214 or 66%) compared to region B (110 or 34%).
Almost half of their properties are commercial (46%) followed by retail (37%) and then industrial (17%).
Of all the prpoperties in the portfolio, the majority are commercial properties in region A (104 or 32%).
followed by retail properties in region A (70 or 22%).
The smallest component of their property portfolio consists of industrial properties in region B (only 16 or 5%).
(e) Portfolio performance :
Net profit % across the entire portfolio is normally distributed (histogram) with an average return of 8.6% and a
standard deviation of 3/6%. NP% ranged from the lowest of -4.2% to the highest of 20.3%.
From the cumulative % distribution, 75% of all properties (cumulative 86.1% - cumulative 11.4%) earned a NP% of
between 5% and 12.5% p.a.
Region B (9.8%) has outperformed region A (7.9%) by almost 2% on average, while commercial (9.0%) and retail (9.5%)
have significantly outperformed industrial properties (5.4%).
The worst performing segment is industrial properties in region A (4.9%) while the best performing properties are
commercial properties in region B (12.3%).
There are 15 (4.6%) properties that are under-performing (with less than a 5% net profit % p.a.).(see histogram).
Volatility of NP% is fairly consistent across the segments (approx. 3.5%), except for higher variability noted in the
industrial properties of region B (4.2%).
Growth potential (high NP% p.a. segments) is mainly in commercial properties in region B which represents only 14% of the current portfolio
retail properties in region A (currently only constitute 22% of the current portfolio).

(f) Recommendations: Dispose of the worst performing industrial properties in both regions A and B and
purchase more commercial properties in region B followed by retail properties in region A.
CHAPTER 3

EXPLORATORY DATA ANALYSIS - DESCRIBING DATA

NUMERIC DESCRIPTIVE STATISTICS

Exercise 3.1 (a) median (b) mode (c) mean

Exercise 3.2 Upper quartile

Exercise 3.3 Statements (c) and (f). The mode would be more appropriate (both are categorical)

Exercise 3.4 (a) False (b) False (c) True (d) False (e) False
The new median mass will depend only on the masses of parcels in the 3rd and 4th
ordered positions out of the 6 positions (after adding the extra parcel).
Also, the rank order position of this extra parcel's mass is unknown. It could be the
4th, 5th or 6th ranked mass, but this depends on the masses of parcels that are heavier
than the current median mass.

If the 4th ranked mass is also equal to 6.5 kg, then the new median will not increase.
If, on the other hand, the 4th ranked mass is greater than 6.5 kg, then the median will
increase. Therefore the only statement that can be made with complete certainty is (c),
(i.e. that it is impossible for the new median mass to be less than it was.)

Exercise 3.5 Correct method is (b). Use the formula for the arithmetic mean (Formula 3.1)
By definition, Mean = Σx / n

Given Mean = 4.1 and n = 9245, it is possible to


compute Σx (total number of persons) = Mean x n .
i.e. total no. persons in Mossel Bay = 4.1 x 9245 = 37905 (rounded)
Exercise 3.6 File: X3.6 - equity returns.xls
General Equity Unit Trust % Returns

% Returns Deviation Deviation2


9.2 -0.1 0.01
8.4 -0.9 0.81
10.2 0.9 0.81
9.6 0.3 0.09
8.9 -0.4 0.16
10.5 1.2 1.44
8.3 -1 1
Sum 65.1 4.32
n 7

Using Excel

Mean 65.1/7 = 9.3 '=AVERAGE(9.2,...,8.3)

Std Dev √[4.32/(7-1)] = 0.8485 '=STDEV(9.2,...,8.3)


Exercise 3.7 File: X3.7 - luggage weights.xls

Mass (kg) Deviation Deviation2


11 0.43 0.1849
12 1.43 2.0449
8 -2.57 6.6049
10 -0.57 0.3249
13 2.43 5.9049
11 0.43 0.1849
9 -1.57 2.4649
Sum 74 17.7143
n 7
Using Excel

(a) Mean 74/7 = 10.57 '=AVERAGE(11,...,9)

Std Dev √[17.7143/(7-1)] = 1.7183 '=STDEV(11,...,9)

(b) On average, each passenger's hand luggage weighs 10.57 kg.

68.3% of all hand luggage is likely to weigh between 8.85 kg and 12.29 kg.
(This corresponds to one standard deviation limits about the mean).

(c) Coefficient of Variation 1.7183/10.57% = 16.26%

(d) The variation in hand luggage weights is moderate (close together).


Exercise 3.8 File: X3.8 - bicycle sales.xls

Bicycles sold Deviation Deviation2 Sorted data


25 1.4 1.96 16
18 -5.6 31.36 18
30 6.4 40.96 18
36 12.4 153.76 19
18 -5.6 31.36 20
20 -3.6 12.96 24
16 -7.6 57.76 25
24 0.4 0.16 30
30 6.4 40.96 30
19 -4.6 21.16 36
Sum 236 392.4
n 10

(a) Mean = 236/10 = 23.6


On average, 23.6 bicycles are sold each month.

Median = (20 + 24)/2 = 22 (Median sales lies in the 5.5th position)


For half of the months (i.e. 5 months), bicycles sales were
less than 22 bicycles per month.

(b) Range = 36 - 16 = 20
The range of sales between the worst and best months was 20 bicycles.
(i.e. In the worst sales month, 16 were sold; in the best month, 36 were sold.

Variance = 392.4/(10-1) = 43.6

Standard deviation = √43.6 = 6.603


68.3% of all monthly bicycle sales are likely to lie between 17 and 30.2.

th
(c) Lower Quartile (Q1) = 18 (2.5 position)
25% of monthly bicycle sales were less than or equal to 18.
Or: No more than 18 bicycles per month were sold in 25% of the months.
Note: Using Excel : QUARTILE(data values,1) = 18.25

Upper Quartile (Q3) = 27.5 (7.5th position)


25% of monthly bicycle sales were above 27.5.
Or: More than 28 (27.5) bicycles per month were sold in 25% of the months.
Note: Using Excel : QUARTILE(data values,3) = 28.75

(d) Approximate Skewness = 3x(23.6 - 22)/6.603 = 0.7269


There is moderate positive skewness in monthly bicycle sales.
(i.e. There are one / two months with relative high bicycle sales)
(e) Box plot of monthly bicycle sales

Interpretation
Monthly bicycle sales range between 16 and 36.
The median monthly sales is 22. The positive skewness shows a wider spread
of monthly sales toward the months of high sales.

(f) Opening monthly stock level = 23.6 + 6.603 = 30.2 bicycles in stock
If orders = 30, then the dealer will have sufficient bicycle stock to meet demand.
Exercise 3.9 File: X3.9 - setting times.xls

Setting time Deviation Deviation2


27 3 9
18 -6 36
21 -3 9
22 -2 4
20 -4 16
28 4 16
31 7 49
25 1 1
24 0 0
Sum 216 140
n 9

(a) Mean = 216/9 = 24 minutes

Std dev = √140/(9-1) = 4.183 minutes

(b) Coefficient of Variation = 4.183/24 = 17.43 %

(c) No, since the consistency index is greater than 10%, this consignment
will not be approved for dispatch to the clients.
Exercise 3.10 File: X3.10 - wage increases.xls

% Increases Deviation Deviation2 Sorted %


5.6 -0.83 0.6910 3.4
7.3 0.87 0.7547 4.8
4.8 -1.63 2.6610 5.3
6.3 -0.13 0.0172 5.6
8.4 1.97 3.8760 5.8
3.4 -3.03 9.1885 5.8
7.2 0.77 0.5910 5.8
5.8 -0.63 0.3985 6.2
8.8 2.37 5.6110 6.3
6.2 -0.23 0.0535 7.2
7.2 0.77 0.5910 7.2
5.8 -0.63 0.3985 7.3
7.6 1.17 1.3660 7.4
7.4 0.97 0.9385 7.6
5.3 -1.13 1.2797 8.4
5.8 -0.63 0.3985 8.8
Sum 102.9 28.8144
n 16

Using manual computations

(a) Mean = 102.9/16 6.43 %


Median = (6.2+6.3)/2 = 6.25 % (lies in the 8.5th position)

(b) Variance = [28.8144/(16-1)] = 1.921


Std dev = √28.8144/(16-1) = 1.386 %

(c) Lower limit = 6.43-2*(1.386) 3.66


Upper limit = 6.43+2*(1.386) 9.20
95.5% of all agreed wage increases lie between 3.66% and 9.2%.

(d) CV = (1.386/6.43) % = 21.56 %


Agreed wage increases are only moderately consistent.
Using Excel
Excel 's Data Analysis option Excel 's Function Keys

Wage increases
(a) Mean 6.43 '=AVERAGE(data values)
Standard Error 0.346
Median 6.25 '=MEDIAN(data values)
Mode 5.8
(b) Standard Deviation 1.386 '=STDEV(data values)
Sample Variance 1.921 '=VAR(data values)
Kurtosis 0.196
Skewness -0.286
Range 5.4
Minimum 3.4
Maximum 8.8
Sum 102.9
Count 16

(c) and (d) must be computed manually.


Exercise 3.11 Bank Trainee Exam Performance

Group 1 Group 2
Mean 76 64
Variance 110 88
Sample size 34 26

Std deviation √110 = √88 =


10.488 9.381

Group 1 Group 2
Coefficient (10.488/76)% (9.381/64)%
of Variation = 13.8 14.66

Interpretation
Both groups performed consistently well. The difference in CV% measures is marginal.
However, group 1 trainees performed marginally more consistently than group 2 trainees.
Exercise 3.12 File: X3.12 - meal values.xls

(a) Random variable - value of a restaurant meal (in Rand)


Data type - numerical, continuous, ratio-scaled (Using Data Analysis in Excel )

Meal values
(b) Mean 1119/20 = = R55.95 Mean 55.95
Standard Error 3.20
∑(deviation)² = 3902.95 Median 53
n= 20 Mode 44
Standard Deviation 14.33
Variance = 3902.95/(20-1) = 205.42 Sample Variance 205.42
Std deviation = √205.42 = 14.33 Kurtosis 0.26
Skewness 0.74
Range 55
(c) Median Minimum 35
th th
Average the Rand values in 10 and 11 positions. Maximum 90
= (51+55)/2 = R53 Sum 1119
Half off the meals were valued at R53 or less. Count 20

Ranked Meal
Position Value
1 35
2 36
3 44
4 44
5 44
6 47
7 48
8 48
9 50
10 51 Median is midpoint between
11 55 these two middle values (R51 and R55)
12 56
13 58
14 62
15 65
16 65
17 69
18 72
19 80
20 90

(d) Mode R44 occurs 3 times (see grouped ranked values in (c) above).

(e) There is moderate positive skewness caused by two high meal values (i.e. R80 and R90).
Hence recommend the median as the most representative central location meal value.
Exercise 3.13 File: X3.13 - days absent.xls

(a) Mean = 237/23 10.3 days absent (Using Data Analysis in Excel )

Median Median is found in the (23+1)/2 th position days absent


i.e. 12th position Mean 10.30
Standard Error 1.33
Median = 9 days absent Median 9
Mode 5
Position Days absent Standard Deviation 6.36
1 2 Sample Variance 40.49
2 4 Skewness 1.38
3 4 Range 28
4 5 Minimum 2
5 5 Q1 position Maximum 30
6 5 and value Sum 237
7 6 Count 23
8 6
9 6 Q1 5.5
10 8 Q3 15
11 9
12 9 Median position and value
13 10
14 10
15 10
16 12
17 15 Q3 position
18 15 and value
19 15
20 16
21 17
22 18
23 30

Mode There are 4 possible modal values (5; 6; 10 and 15 days).


All occur with a frequency of 4.
Note: Only the first modal value is reported in Excel .
The mode is an unreliable measure of central location in this study.

Interpretation
On average, an employee is absent for 10.3 days over this 9-month period.
Half the employees were absent for up to 9 days.
The most common number of days absent was 5 (or 6, or 10 or 15) days

(b) Lower Quartile (approximated manually)


Q1 position = (23/4) = 5.75th position
Q1 value in this position is = 5+(0.75*(5 - 5) = 5 days absent
Using Excel 5.5 days absent =QUARTILE(data range,1)
25% of employees were absent for no more than 5 (or 5.5) days altogether.
Upper Quartile (approximated manually)
Q3 position = (23*3/4) = 17.25th position
Q3 value in this position is = 15+(0.25*(15 - 15) = 15 days absent
Using Excel 15 days absent =QUARTILE(data range,3)
25% of employees were absent for more than 15 days over this 9-month period.

(c) Average per 9-month = 10.3 days


Average per 1-month = 1.14 days absent per month = (10.3 / 9) = ave per month
Since the monthly average is above 1 day (actually 1.14 days), the company is
not succesfully managing its absenteeism levels.
Exercise 3.14 File: X3.14 - bad debts.xls

(a) Average = 79.3/17 4.665% (Using Data Analysis in Excel )

∑(deviation)² = 54.85882 bad debts %


n= 17 Mean 4.665
Standard Error 0.45
Variance = 54.85882/(17-1) = 3.43 Median 5.4
Std deviation = √(3.43) = 1.85 Mode 2.2
Standard Deviation 1.85
Sample Variance 3.43
th
(b) Median Median is found in the (17+1)/2 position Kurtosis -1.49
th
i.e. 9 position Skewness -0.35
Median = 5.40% Range 5.4
Minimum 1.8
Rank Ordered Maximum 7.2
Position bad debts % Sum 79.3
1 1.8 Count 17
2 2.2
3 2.2 Q1 2.6
4 2.4 Lower Quartile position Q3 6.1
5 2.6 and value
6 3.4
7 4.4
8 4.7
9 5.4 Median position and value
10 5.7
11 5.7
12 5.8 Upper Quartile position
13 6.1 and value
14 6.3
15 6.6
16 6.8
17 7.2

(c) Average: On average, each furniture retailer has a bad debt % of 4.665%.
Median: 50% of furniture retailers have a bad debt % of 5.4% or less.
Since the mean < median, there is evidence of negative skewness.
Hence propose the use of the median as the representative central value.

(d) There are two modal values (2.2% and 5.7%) both occurring with frequency of 2.
This makes the mode an unreliable measure of central location.

(e) Skewness coefficient Values required for Formula 3.14


(Formula 3.14) ∑(x - x(bar))3 = -31.2227
n= 17
s (std dev) = 1.85

3
Then Skp = (17*(-31.2227))/((17-1)*(17-2)*1.85 ) = -0.35

Since the skewness coefficient is close to zero, there is only


There is evidence of moderate negative skewness in the data on bad debts %
(i.e. only a few fairly low bad debt % values - the majority are higher).
(f) Lower Quartile (approximated manually)
Q1 position = (17/4) = 4.25th position
Q1 value in this position is = (2.4*+(0.25*(2.6 - 2.4)) = 2.45%
Using Excel = 2.6% =QUARTILE(data range,1)
25% of furniture retailers have a bad debts % of no more than 2.45% (or 2.6%).

Upper Quartile (approximated manually)


Q3 position = (17*3/4) = 12.75th position
Q3 value in this position is = (5.8*+(0.75*(6.1 - 5.8)) = 6.025%
Using Excel = 6.1% =QUARTILE(data range,3)
25% of furniture retailers have a bad debts % of more than 6.025% (or 6.1%).

(g) The average % bad debts is 4.665% while the median % bad debts is 5.4%. Since
there is moderate negative skewness (Skp = -0.35), the median should be used
as the representative central value. Thus, since the median % of bad debts,
is above 5% (median = 5.4%), an advisory note should be sent out.
Exercise 3.15 File: X3.15 - fish shop.xls

(a) Average T/O bins midpoint (x ) freq (f ) xf


(or Mean) 500 - 750 625 15 9375
750 - 1000 875 23 20125
1000 - 1250 1125 55 61875
1250 - 1500 1375 92 126500
1500 - 1750 1625 65 105625
1750 - 2000 1875 50 93750
Total 300 417250

(Formula 3.5) Average = 417250/300 R 1,390.80

Interpretation The average daily turnover for the fish shop is R1390.80 per day.

(b) Median T/O bins freq (f ) ∑f


500 - 750 15 15
750 - 1000 23 38
1000 - 1250 55 93 Q1 interval
1250 - 1500 92 185 Median Interval
1500 - 1750 65 250 Q3 interval
1750 - 2000 50 300
Total 300

Median position = (300/2) = 150th position out of 300 positions


Median value therefore lies in the 4th interval [i.e. Between R1250 and R1500]

Median = 1250+(250*(150-93)/(185-93)) R 1,404.90


(based on Formula 3.2)

(c) Mode Modal value lies in 4th interval [1250 - 1500]


as it has the highest frequency count of 92 days

Mode = 1250+(250*(92-55)/(2*92-55-65)) = R 1,394.50


(based on Formula 3.3)

(d) Lower Quartile (using Formula 3.7)


Q1 position = (300/4) = 75th position which lies in the interval [1000 - 1250]
Q1 value in this position is = (1000+250*(75-38)/(93-38)) = R 1,168.18
The maximum turnover for the slowest 25% of trading days was R1168.18.

(e) Upper Quartile (using Formula 3.8)


Q3 position = (300*3/4) = 225th position which lies in the interval [1500 - 1750]
Q3 value in this position is = (1500+250*(225-185)/(250-185)) = R 1,653.85
A minimum turnover of R1653.85 was generated on the 25% of busiest trading days.
Exercise 3.16 File: X3.16 - grocery spend.xls

(a) Average % Spend midpoint (x ) freq (f ) xf


(Mean) 10 - 20 15 6 90
20 - 30 25 14 350
30 - 40 35 16 560
40 - 50 45 10 450
50 - 60 55 4 220
Total 50 1670

(Formula 3.5) Average = 1670/50 33.4%

Interpretation On average, a family spends 33.4% of their income on groceries.

(b) (i) Median % Spend freq (f ) ∑f


10 - 20 6 6
20 - 30 14 20 Q1 interval
30 - 40 16 36 Median Interval
40 - 50 10 46 Q3 interval
50 - 60 4 50
Total 50

Median position = (50/2) = 25th


Median value therefore lies in the 3rd interval [30 - 40]

(Formula 3.2) Median = 30+(10*(25-20)/(36-20)) 33.125%

Interpretation 50% of families spend no more than 33.125% of their income on groceries.

(b) (ii) Lower Quartile (Q1) (using Formula 3.7)


Q1 position = (50/4) = 12.5th position which lies in the interval [20 - 30]
Q1 value in this position is = 20+(10*(12.5-6)/(20-6)) = 24.64%

Interpretation 25% of families spend no more than 24.64% of their income on groceries.

(c) Upper Quartile (Q3) (using Formula 3.8)


Q3 position = (50*3/4) = 37.5th position which lies in the interval [40 - 50]
Q3 value in this position is = 40+(10*(37.5-36)/(46-36)) = 41.5%

Interpretation 25% of families spend more than 41.5% of their income on groceries.
Exercise 3.17 File: X3.17 - equity portfolio.xls

Shares Price Value


40 15 600
10 20 200
5 40 200
50 10 500
105 Total 1500

Average value / equity = 1500/105 R 14.29


(This is a weighted average measure)
(Using Formula 3.5)
Exercise 3.18 File: X3.18 - car sales.xls

Cars sold Price Value


5 25000 125000
12 34000 408000
3 55000 165000
20 Total 698000

Average Price / Car = 698000/20 R34 900


(This is a weighted average measure)
(Using Formula 3.5)
Exercise 3.19 File: X3.19 - rental increases.xls

(Use Formula 3.4)


4
Geometric mean = √(1.16*1.14*1.1*1.08) - 1 = 0.11955

Interpretation
The average annual % escalation rate in office rentals is 11.955%

Using Excel =GEOMEAN(1.16,1,14,1,1,1,08) - 1 = 0.11955


Exercise 3.20 File: X3.20 - sugar increases.xls

(a) (Using Formula 3.4)


6
Geometric mean = √(1,05*1,12*1,06*1,04*1,09*1,03) - 1 = 0.06456

Interpretation
The average annual % increase in the sugar price (per kg) has been 6.456%

Using Excel =GEOMEAN(1.05,1.12,1.06,1.04,1.09,1.03) - 1 = 0.06456

(b) The geometric mean is more appropriate since the base value of each
percentage change is different .
Each year's percentage change is based on the previous year's sugar price.
Exercise 3.21 File: X3.21 - water usage.xls

Using Excel 's Descriptive Statistics option in Data Analysis

water usage
(a) Mean 21.2
Standard Error 1.751
Median 19.5
Mode 25
(b) Standard Deviation 9.590
Sample Variance 91.959
Kurtosis 1.744
Skewness 1.165
Range 42
Minimum 8
Maximum 50
Sum 636
Count 30

(c) Using Excel 's Function Key - QUARTILE

Lower Quartile - Q1 15 '=QUARTILE(data range,1)


Upper Quartile - Q3 25.75 '=QUARTILE(data range,3)

(d) Interpretation - central location measures


The average monthly water consumption per household is 21.2 kl.
50% of households consume no more than 19.5 kl per month.
The most frequently occuring monthly consumption is 25 kl (Note that this modal
(value is a misleading and misrepresentative measure at it only occurs 3 times).

Interpretation - dispersion measures


Based on the mean and standard deviation values, approximately 95.5% of
households consume between 2.02 kl and 40.38 kl per month.
Note, that the data is heavily skewed to the right (see skewness = 1.165), implying
that there are a few households that consume high volumes of water per month.
This skewness makes the interpretation given above unreliable as both the mean
and the standard deviation are likely to be inflated (or over-estimated) due to the
presence of a few high-valued outliers).

Interpretation - quartiles
25% of households consume at most 15 kl of water per month.
The 25% of heavy water consumers use at least 25.75 kl per month.

(e) Households in Paarl suburb 750


- expected total usage per month 21.2 x 750 = 15 900 kl
- expected total usage per year 15900 x 12 = 190 800 kl
Exercise 3.22 File: X3.22 - veal dishes.xls

(a) Random variable - cost of a veal cordon bleu meal at a Durban restaurant
Data type - numerical, continuous, ratio-scaled.

(b) Using Excel 's Descriptive Statistics option in Data Analysis

cordon bleu meal price


Mean 61.25
Standard Error 1.99
Median 59
Mode 48
Standard Deviation 10.54
Sample Variance 111.08
Kurtosis 0.62
Skewness 0.78
Range 45
Minimum 45
Maximum 90
Sum 1715
Count 28

Interpretation
On average, a patron can expect to pay R61.25 for a veal cordon bleu meal
at a Durban restaurant.
50% of Durban restaurants charge no more than R59 for a veal cordon bleu meal.

(c) There are two modal values (R48 and R55). Both occur with a frequency of 3.
It is a misleading value because of its low frequency of occurrence.
Note: Excel only shows the first occurrence of a modal value (i.e. R48) in its output.

(d) Standard deviation = R 10.54


This means that 68.3% of Durban restaurants are likely to charge between
R50.71 (61.25 - 10.54) and R71.79 (61.25 + 10.54) for a veal cordon bleu meal.

(e) Skewness Coefficient = 0.78 Formula 3.14


Skewness coefficient (approx) = 0.64 Formula 3.15
The relative high positive skewness is caused by two Durban restaurants
charging high prices (R80 and R90) for a veal cordon bleu meal.

(f) Both measures indicate that there is moderate-to-high positive skewness, hence
the median cost of R59 would be a more representative measure of central location.

(g) Upper Quartile (Q3) R68 =QUARTILE(data range,3)


25% of Durban restaurants charge at least R68 for a veal cordon bleu meal.

(h) Lower Quartile (Q1) R 54.75 =QUARTILE(data range,1)


25% of Durban restaurants charge no more than R54.75 for a veal cordon bleu meal.

(i) 90th percentile value R 72.90 =PERCENTILE(data range,0.9)


Exercise 3.23 File: X3.23 - fuel bills.xls

(a) Using Excel 's Descriptive Statistics option in Data Analysis

fuel bill
(i) Mean Mean 418
Standard Error 13.128
(ii) Median Median 398
Mode 350
(iv) Standard deviation Standard Deviation 113.693
(iii) Variance Sample Variance 12926.054
Kurtosis -0.503
(v) Skewness Skewness 0.601
Range 420
Minimum 256
Maximum 676
Sum 31350
Count 75

(b) Interpretation - central location


The average monthly fuel bill per motorist is R418; while half of the motorists
spend no more than R398 on their monthly fuel.

Interpretation - standard deviation


68.3% of monthly fuel bills range between R304.31 and R531.69 (one std dev from mean)
95,5% of monthly fuel bills range between R190.61 and R645.39 (within 2 std devs of mean)

(c) Interpretation - skewness


There is moderate positive skewness (Skp = 0.601) caused by 8 motorists who spend
more than R600 per month - well above the majority of motorists' fuel spend.

(d) Coefficient of Variation CV% 113,693/418 = 27.2%


There is only moderate consistency across monthly fuel bills of motorists
This greater relative spread could be caused by different size vehicles and varying distances travelled.

(e) Lower Quartile - Q1 R332.5 =QUARTILE(data range,1)


Upper Quartile - Q3 R502.5 =QUARTILE(data range,3)
Inter Quartile Range = (Q3 - Q1) R170
The middle 50% of monthly fuel bills span R170 from a low of R332.50 to a high of R502.50.

(f) Five-Number Summary table Using the Excel Function Keys

Minimum R256 =MIN(data range) or =QUARTILE(data range,0)


Lower Quartile Q1 R332.5 =QUARTILE(data range,1)
Median R398 =MEDIAN(data range) or =QUARTILE(data range,2)
Upper Quartile Q3 R502.5 =QUARTILE(data range,3)
Maximum R676 =MAX(data range) or =QUARTILE(data range,4)
(g) Box plot

(h) Interpretation of Box Plot


There is clear evidence of moderate positive skewness (skewed-to-the-right) in fuel bills.
There are a few motorists who spend a large amount on fuel a month.

(i) Total amount of fuel consumed monthly by Paarl motorists for commuting to work:

Expected total monthly fuel bill = R418 * 25000 R 10,450,000


(average bill / motorist x no.motorists)

Expected Total fuel consumed (in litres) = R10450000/10 = 1 045 000 litres
(Expected total monthly cost / cost per litre)
Exercise 3.24 File: X3.24 - service periods.xls

(a) Using Excel 's Descriptive Statistics option in Data Analysis

service periods
(i) Mean Mean 7
Standard Error 0.384
(ii) Median Median 6
Mode 6
(iii) Std Dev Standard Deviation 3.838
Sample Variance 14.727
Kurtosis -0.437
(iv) Skewness Skewness 0.553
Range 15
Minimum 1
Maximum 16
Sum 700
Count 100

(b) Interpretation - central location


On average, each engineer has 7 years of service with a company
Half of all engineers spend no more than 6 years with a company

Interpretation - standard deviation


68.3% of engineers spend between 3.16 and 10.84 years with a company.
Similar intervals can be computed for 2 and 3 std devs from the mean.
Note: when the lower limit is computed to be negative, in practice it is zero.

Interpretation - skewness
There is a moderate positive skewness (Skp = 0.553) in length of service periods,
meaning that a few engineers have long service periods with their company.

(c) Using Excel 's Histogram option in Data Analysis

Frequency Distribution - Years of Service

Years of Service Count


0-3 21
3-6 30
6-9 24
9 - 12 15
12 - 15 8
15 - 18 2
(d) Lower limit = 7 - 3.838 = 3,16 years
Upper limit = 7 + 3.838 = 10,84 years
68.3% of engineers spend between 3.16 years and 10.84 years with a company.

(e) Percent of members with less than 3 years of service


Cumulative % up to 3 years = 21/100 = 21%

Percent of members with more than 12 years of service


Cumulative % above 12 years = 10/100 = 10%

They meet the guidelines for "new blood" (21,2%) ,but have
far fewer "experienced" members (only 10%) than their guidelines require.
Exercise 3.25 File: X3.25 - dividend yields.xls

(a) Random variable dividend yield (%) of a company


Data type numeric, ratio-scaled

(b) Using Excel 's Descriptive Statistics option in Data Analysis

Dividend yields
(i) Mean Mean 4.273
Standard Error 0.218
(ii) Median Median 4.1
(iii) Mode Mode 2.8
(iv) Std Dev Standard Deviation 1.444
Sample Variance 2.084
Kurtosis -0.359
(v) Skewness Skewness 0.259
Range 6.1
Minimum 1.5
Maximum 7.6
Sum 188
Count 44

(c) Mean The average dividend yield per company is 4.273%

Median Half of the dividend yields are at or below 4.1%

Mode A misleading measure because there are 4 dividend yield values


(viz. 2,8, 3,6; 4,1; 5,1) of equal modal frequency (of 3)

Std dev 68.3% of all dividend yields lie between 2.83% and 5.72%
Similarly for 2 and 3 std devs from the mean

Skewness There is very slight positive (right) skewness (Skp = 0.259).


The histogram can be assumed to be normally distributed.

(d) the Mean (Average) as there is minimal skewness present in the data.

(e) Bin yields Frequency


Below 2% 3
2 - 3.5% 11
3.5 - 5% 16
5 - 6.5% 11
6.5 - 8% 3
(f) Five-Number Summary Table
Minimum 1.5 =QUARTILE(data range,0)
Lower Quartile Q1 3.175 =QUARTILE(data range,1)
Median 4.1 =QUARTILE(data range,2)
Upper Quartile Q3 5.15 =QUARTILE(data range,3)
Maximum 7.6 =QUARTILE(data range,4)

(g) Box plot Dividend Yield %

The very slight positive skewness can be seen from the longer tail
on the right of the box plot. It implies a few companies have achieved
significantly higher dividend yields than the majority of the sample.

(h) Minimum dividend yield achieved by the top performing 10% of companies
6.17% =PERCENTILE(data range,0.90)
The top 10% of companies achieved at least a dividend yield of 6.17%

(i) % of companies who did not declare more than a 3.5% dividend yield
Cumulative frequency up to upper limit of 3.5 = (3 + 11) = 14.
% Cumulative = 14 / 44 % = 31.8%
Almost 1/3 (31.8%) of companies did not declare more than a 3.5% dividend yield.
Exercise 3.26 File: rosebuds.xls

(a) Random variable - the unit price of a rosebud (in cents)


Data type - numerical, continuous, ratio-scaled

(b) Using Excel 's Descriptive Statistics option in Data Analysis

selling price
(i) Mean 60.312
Standard Error 0.319
(iii) Median 59.95
Mode 60.6
(ii) Standard Deviation 3.188
Sample Variance 10.160
Kurtosis 1.770
(iv) Skewness 0.932
Range 18.3
Minimum 55.2
Maximum 73.5
Sum 6031.2
Count 100

Interpretation
Mean: The average selling price of rosebuds is 60.312 cents.
Median: 50% of rosebuds sold for no more than 59.95c.
Std dev: 68.3% of rosebud unit prices are likely to lie between 57.12c and 63.5c.
Skewness: There is excessive positive skewness (one very high unit price)

(c) Coefficient of Variation CV% 3.188/60.312% = 5.3%


There is very low variability amongst unit selling prices of rose buds

(d) Lower Quartile Q1 57.7 cents =QUARTILE(data range,1)


Upper Quartile Q3 62.45 cents =QUARTILE(data range,3)

(e) The highest unit selling price of the cheapest 25% of sales is 57.7 cents (i.e. Q1)

(f) The minimum unit selling price of the highest-priced 25% of sales is 62.45 cents (i.e. Q3)
Overall interpretation of (a), (d), (e) and (f)
Unit selling prices ranged from 55.2c to 73.5c where 50% of the selling prices
were above 59.95c. 25% of sales were below 57.7c while the most expensive 25%
of unit selling prices were above 62.45c.

(g) 90th percentile 64.6c =PERCENTILE(data range,0.9)

(h) 10th percentile 56.6c =PERCENTILE(data range,0.1)

(i) Five-Number Summary Table and Boxplot

Minimum 55.2 =QUARTILE(data range,0)


Lower Quartile Q1 57.7 =QUARTILE(data range,1)
Median 59.95 =QUARTILE(data range,2)
Upper Quartile Q3 62.45 =QUARTILE(data range,3)
Maximum 73.5 =QUARTILE(data range,4)
Boxplot

Interpretation
The distribution of unit selling prices of rosebuds is skewed to the right.
There is also one extremely high unit selling price of 73.5c which can be
considered an outlier.
Overall, unit selling prices of rosebuds ranged between 55.2c and 73.5c.
25% of unit selling prices lay below 57.7 c (i.e. lower quartile); while
the middle 50% of unit selling prices ranged from 57.7c to 62.45c.
The "best" 25% of unit selling prices achieved were above 62.45c (Q3).
Finally, the median (middle) unit selling price achieved was 59.95c.
Mini Case Study 3.27 File: X3.27 - savings balances.xlsx

1 (a) Frequency Distribution

Savings Intervals Frequency Histogram of Savings Balances (R10's)


0 - ≤ 200 21
80
201 - ≤ 400 71 71
401 - ≤ 600 56 70
601 - ≤ 800 19 60 56

Number of Clients
801 - ≤1000 4
More 50
4
40
30
21 19
20
10 4 4
0
0 - ≤ 200 201 - ≤ 400 401 - ≤ 600 601 - ≤ 800 801 - ≤1000 More
Savings Intervals

1 (b) Descriptive Statistics

Savings Balance (R10's)


Mean 421.86 Lower Quartile 266.5
Standard Error 16.247 Upper Quartile 520.5
Median 385
Mode 326
Standard Deviation 214.93
Sample Variance 46195.1765
Kurtosis 4.727
Skewness 1.634
Range 1383
Minimum 85
Maximum 1468
Sum 73826
Count 175

1 (c ) Married Single Grand Total


Female 23 41 64
Male 83 28 111
Grand Total 106 69 175

1 (d) Married Single Grand Total


Female 512.6 563.7 545.3
Male 368.0 299.3 350.7
Grand Total 399.4 456.4 421.9

2 (a) Gender Categorical Nominal-scaled


Marital status Categorical Nominal-scaled
Month-end balances Numeric Ratio-scaled

2 (b) Refer to 1 (a) (Histogram) and 1 (b) (Descriptive Statistics)

The average savings balance of bank clients is R421.86.


50% of these clients have month-end balances of less than R385 (median).
From the histogram, 21 clients have month-end balances of less than R200 and 8 are above R800.
The 8 banking clients with relatively high month-end savings balances (skewness = 1.643 > 1.0) skew the average
towards these higher balances and distorts the overall picture.
Therefore the median month-end balance of R385 is a more representative indicator of savings balances.
25% of their clients have month-end balances of less than R266.6, while the top 25% of savers
have month-end balances in excess R520.5 with 4 clients above R1000 (maximum = R1468).

Refer to 1 (c) (Count of Gender and Marital Status)

Amongst the female clients, the majority (64%) are single (41/64), while amongst the male clients,
the majority (75%) are married.
Thus they are attracting mainly single females; and married males.

Refer to 1 (d) (Breakdown Table of Average Savings by Gender and Marital Status)

Single females are saving the most (average of R563,7) while single males are the worst savers with
an average savings balance of only R299.3 (compared to the overall average of R421.9).
Also females save more, on average (R545) than males (R350).
Single bank clients save more, on average (R456.4) than married bank clients (R399.4).
2 (c) Plan of Action by Bank
Bank should target females in general (as they comprise only 37%) of the current client base
but have the largest average balances of all clients.
Single females (23%) in particular should be targetted to attract more high savers to the bank.
They should encourage their main client base - married males - to save more.
Mini Case Study 3.28 File: X3.28 - medical claims.xlsx

1 (a) Frequency Distribution

Claims Ratio Count % Histogram of Claims Ratios


Below 0.1 6 4.0
40 35 36
0.1 - 0.5 35 23.3
35 30
0.5 - 0.9 26 17.3

No. of Members
30 26
0.9 - 1.3 36 24.0
25
1.3 - 1.7 30 20.0
20
1.7 - 2.1 14 9.3 14
15
2.1 - 2.5 2 1.3 10 6
Above 2.5 1 0.7 5 2 1
150 100 0

Claims Ratio Intervals

1 (b) Descriptive Statistics

Claims Ratio
Mean 0.974 Lower Quartile 0.460
Median 0.996 Upper Quartile 1.421
Mode #N/A
Standard Deviation 0.594
Sample Variance 0.353
Skewness 0.242
Range 2.672
Minimum 0.013
Maximum 2.685
Sum 146.1
Count 150

1(c) and (d) Cross-tabulation Table / Breakdown table

Age
Marital 26 - 35 36 - 45 46 - 55 Row totals
Married average 1.326 0.915 1.123 1.112
std dev 0.552 0.589 0.629 0.610
minimum 0.147 0.051 0.013 0.013
maximum 2.685 1.989 2.408 2.685
count 23 27 35 85
Single average 0.919 0.474 0.947 0.793
std dev 0.543 0.356 0.500 0.524
minimum 0.014 0.024 0.218 0.014
maximum 1.905 1.262 1.814 1.905
count 36 19 10 65
average 1.078 0.733 1.084 0.974
Column totals std dev 0.577 0.547 0.602 0.594
minimum 0.014 0.024 0.013 0.013
maximum 2.685 1.989 2.408 2.685
count 59 46 45 150

2 (a) Marital Status Qualitative / Categoric Nominal


Age Group (bands) Qualitative / Categoric Ordinal
Claims Ratio Quantitative / Numeric Ratio, Continuous

2 (b) Claims Ratio Pattern of Members


The average claims ratio is 0.974. This means that, on average, members are claiming
as much as they contribute. The median claims ratio is a similar value (0.996).
This means that at least half the members are claiming more than they contribute.
The distribution (see histogram) appears to be bi-modal showing a low claims ratio (i.e. less than half their contributions)
by at least a quarter of members (Q1 = 0.46) and a high claims ratio by at least half of the
members (50% greater than median claims ratio of 0.996). At least 25% of members are
claiming significantly more than they contribute (Q3 = 1.42).
Only 4% of members claim less than 10% of their contributions (first interval of histogram).
Also at least 11% (top three intervals of histogram) are claiming at least twice as much as they contribute.

Refer to 1(c) and (d)


It would appear that married members claim significantly more (1.112) than single members (0.793).
Married members also make up the majority of members (85/150 = 57%).
Also, the younger members (26 - 35) and older members (46 - 55) claim significantly more
(1.078 and 1.084 respectively) than the middle-aged members (36 - 45) (0.733), on average.
These two 'high-claiming' age segments of members represent 69% (104/150) of all members.
The highest claiming segments are the married younger (26-35) (1.326) and married older members (46 - 55) (1.123).
Members from these two segments also have the maximum claims ratio of all members (2.685 and 2.408 respectively).
Collectively these members represent 39% (58/150) of all members of the medical scheme.
The lowest claiming segment is the middle-aged single members (0.474) but they only
comprise 13% (19/150) of all scheme members.
The range of claims ratios is highest amongst married (25-35 years) (0.147 - 2.685) and married (46 - 55 years) (0.013 - 2.408).
It is lowest amongst single (36 - 45 year) members (0.024 - 1.262).
In conclusion:
Younger and Older married members claim more on average - and above their contribution levels in most cases).
They also make up the bulk of membership (57%).
The single members tend to claim less than the married members, but only make up 43% of the membership base.

2 (c) There is cause for concern about the financial viability of the Scheme because:
The claims ratio of married members exceed 1 on average
Within the married group, the age groups 26 -35 and 46 - 55 are claiming more than they contribute
The married group represent the majority of members (57%)
The two age groups within the married group represent 69% of all married (and are claiming more than they contribute).
Overall the scheme is operating close to its breakeven of financial non-viability (overall mean claims ratio = 0.974)
There is significant cross-subsidization of the married younger and older members by mostly
the middle-aged single members.
The mangement of this medical scheme needs to review member contributions from the married younger and older members.
CHAPTER 4

BASIC PROBABILITY CONCEPTS

Exercise 4.1 P(A) = 0.2 means that an event has a 20% chance of occurring.

Exercise 4.2 Mutually exclusive events.

Exercise 4.3 The outcome of one event does not influence / nor is influenced
by / the outcome of the other event.

Exercise 4.4 P(A or B) = P(A) + P(B) - P(A ∩ B) = 0.26 + 0.35 - 0.14 = 0.47

Exercise 4.5 P(X / Y) = P(X ∩ Y) / P(Y) = 0.27 / 0.36 = 0.75


P(Y / X) = P(Y ∩ X) / P(X) = 0.27 / 0.54 = 0.50
Thus P(X / Y) ≠ P(Y / X)
Exercise 4.6 File: X4.6 - economic sectors.xls

(a) Sector Count % count


Mining 45 18.0
Financial 72 28.8
IT 32 12.8
Production 101 40.4
Total 250 100

(b) P(Financial) = 28.8%

(c) P(Not Production) = 100 - 40.4 = 59.6%

(d) P(Mining or IT) = 18 + 12.8 = 30.8%

(e) In (b), the marginal probability was computed.


In (c), used the Complementary probability rule
In (d), used the Addition rule for mutually exclusive events.
Exercise 4.7 File: X4.7 - apple grades.xls

(a) Apple Grades Quantity %


A 795 53.0
B 410 27.3
C 106 7.1
D 189 12.6
Total 1500 100

(b) P(Grade A) = 53%

(c) P(Grade B or D) = 27,3 + 12,6 = 39.9%

(d) P(not (Grade C or D)) = 100 - (7,1 + 12,6) = 80.3%

(e) In (b), the marginal probability was computed.


In (c), used the Addition rule for mutually exclusive events.
In (d), used the Complementary probability rule.
Exercise 4.8 File: X4.8 - employment sectors.xls

(a) Sector Count % count


Formal Business 6678 53.0
Commercial Agriculture 1492 11.8
Subsistence Agriculture 653 5.2
Informal Business 2865 22.7
Domestic Service 914 7.3
Total 12602 100

(b) P(Domestic Service) = 7.3%

(c) P(Commercial or Subsistance Agric) = 11,8 + 5,2 = 17%

(d) P(Informal Business / Business) = 2865/(2865+6678) = 30.02%

(e) In (b), the marginal probability was computed.


In (c), the Addition rule for mutually exclusive events.
In (d), the Conditional Probability rule.
Exercise 4.9 File: X4.9 - qualification levels.xls

(a) Random variable 1 Managerial level (categorical, ordinal-scaled and discrete)


Random variable 2 Qualification level (categorical, ordinal-scaled and discrete)

(b) Managerial Level


Qualification Section Head Dept Head Division Head Total
Matric 28 14 8 50
Diploma 20 24 6 50
Degree 5 10 14 29
Total 53 48 28 129

(c)
(i) P(Matric) = 50/129 = 38.76%
(ii) P(Section head ∩ Degree) = 5/129 = 3.88%
(iii) P(Dept head / Diploma) = 24/50 = 48%
(iv) P(Division head) = 28/129 = 21.71%
(v) P(Division head U Section head) = (53+28)/129 = 62.79%
(vi) P(Matric U Diploma U Degree) = (50+50+29)/129 = 100%
(vii) P(Degree / Dept head) = 10/48 = 20.83%
(viii) P(Division head U Diploma U both) = (28+50-6)/129 = 55.81%

(d) Probability Types and Rules


(i) Marginal probability
(ii) Joint probability and Multiplication rule
(iii) Conditional probability
(iv) Marginal probability
(v) Addition rule for mutually exclusive events
(vi) Collectively exhaustive set of events and Addition rule for mutually exclusive events
(vii) Conditional probability
(viii) Addition rule for non-mutually exclusive events

(e) Yes, since these outcomes cannot occur simultaneously.


Exercise 4.10 File: X4.10 - bonus options.xls

Cash bonus Profit-sharing Share options Total


Admin 28 44 68 140
Production 56 75 29 160
Total 84 119 97 300

(a) P(Cash bonus) = 84/300 = 28.00% Marginal probability


(b) P(Share option) = 97/300 = 32.33% Marginal probability
(c) P(Production ∩ Cash bonus) = 56/300 = 18.67% Joint probability
(d) P(Share option / Admin) = 68/140 = 48.57% Conditional probability
(e) P(Production / Cash bonus) = 56/84 = 66.67% Conditional probability

(f) P(A/B) = P(A) ? P(A/B) = 68/140 = 48.57%


P(A) = 97/300 = 32.33%
Since P(A/B) ≠ P(A), the two events are statistically dependent.
Conclusion: The choice of bonus option and employee work function are associated.

(g) See (a) to (e) above


Exercise 4.11 File: X4.11 - age profile.xls

Department
Age Production Sales Administration Total
<30 60 25 18 103
30-50 70 29 25 124
>50 30 8 35 73
Total 160 62 78 300

(a) (i) P(< 30 years) = 103/300 = 34.33% Marginal probability


(ii) P(Production) = 160/300 = 53.33% Marginal probability
(iii) P(Sales ∩ (30-50) years) = 29/300 = 9.67% Joint probability
(iv) P(>50 / Admin) = 35/78 = 44.87% Conditional probability
(v) P(Production U <30 U both) = (160+103-60)/300 = 67.67% Conditional probability

(b) No, since outcomes of these two events can occur simultaneously.

(c) Let A = >50 and B = Admin.


Is P(A/B) = P(A) ? P(A/B) = 35/78 = 44.87%
P(A) = 73/300 = 24.33%
Since P(A/B) ≠ P(A), the two events are statistically dependent.
Conclusion: There is an association between the Age of an employee and the Department
in which they are employed (e.g. younger employees tend to be in Production).

(d) See (a) above


Exercise 4.12 File: X4.12 - digital cameras.xls

Digital Camera Brand Preference


Usage Canon Nikon Pentax Total
Professional 48 15 27 90
Personal 30 95 65 190
Total 78 110 92 280

(a) P(Professional) = 90/280 = 32.14% Marginal probability


(b) P(Nikon User) = 110/280 = 39.29% Marginal probability
(c) P(Pentax / Personal) = 65/190 = 34.21% Conditional probability

(d) Let A = Professional usage and B = Canon preference.


P(A/B) = P(A) ? P(A/B) = 48/78 = 61.54%
P(A) = 90/280 = 32.14%
Since P(A/B) ≠ P(A), the two events are statistically dependent.
Conclusion: Type of usage and choice of brand are associated.
(e.g. Professionals may prefer Canon while Nikon is favoured by Personal users).

(e) P(Canon ∩ Professional) = 48/280 = 17.14% Joint probability

(f) P(Professional U Nikon U both) = (90+110-15)/280 = 66.07%


Addition rule for non-mutually exclusive events

(g) No, as outcomes of the two events can occur simultaneously as illustrated in (f) above.
Exercise 4.13

(a) Probability Tree

Component A Component B Joint outcomes

(Fail) P(F2) = 0.15 P(F1 ∩ F2) = 0.20 x 0.15 = 0.03

P(F1) = 0.20
(Fail)
(Not fail) P(S2) = 0.85 P(F1 ∩ S2) = 0.20 x 0.85 = 0.17

(Fail) P(F2) = 0.15 P(S1 ∩ F2) = 0.80 x 0.15 = 0.12


(Not fail)
P(S1) = 0.80

(Not fail) P(S2) = 0.85 P(S1 ∩ S2) = 0.80 x 0.85 = 0.68

1.00
Exercise 4.13

(b) P(A) = 0.20 P(B) = 0.15


P(A ∩ B) = P(A) x P(B) = 0.2 * 0.15 = 0.03 (3% chance)

(c) P(Fail) = P(A U B)


P(A U B) = P(A) + P(B) - P(A∩B) = 0.2+0.15-0.03 0.32
Hence P(not Fail) = 1 - P(A U B) = 1 - 0.32 = 0.68 (68% chance)
Exercise 4.14

(a) Probability Tree

Workshop Exam Result Joint outcomes

(Pass) P(S) = 0.80 P(Y ∩ S) = 0.30 x 0.8 = 0.24

P(Y) = 0.30
(Attend)
(Fail) P(F) = 0.20 P(Y ∩ F) = 0.30 x 0.20 = 0.06

(Pass) P(S) = 0.60 P(N ∩ S) = 0.70 x 0.60 = 0.42


(Not attend)
P(N) = 0.70

(Fail) P(F) = 0.40 P(N ∩ F) = 0.70 x 0.40 = 0.28

1.00
Exercise 4.14

(b) Refer to the end nodes of the Probability Tree to answer question 4.14 (b)

(i) P(Pass and Attended workshop) = P(Y ∩ S) = 0.30 x 0.8 = 0.24 24%

(ii) P(Pass) = P(S) = P(S ∩ Y) + P(S ∩ N)

Reading off the Probability Tree P(S ∩ Y) = 0.30 x 0.8 = 0.24


P(S ∩ N) = 0.70 x 0.60 = 0.42

Thus P(S) = 0.24 + 0.42 = 0.66 66%


Exercise 4.15
Using Excel

(i) 6! = 6.5.4.3.2.1 = 720 =FACT(6)


(ii) 3! 5! = (3.2.1)(5.4.3.2.1) = 720 =FACT(3)*FACT(5)
(iii) 4! 2! 3! = (4.3.2.1)(2.1)(3.2.1) = 288 =FACT(4)*FACT(2)*FACT(3)
(iv) 7C4 = 7!/(4!*(7-4)!) = 35 =FACT(7)/(FACT(4)*FACT(3))
(v) 9C6 = 9!/(6!*(9-6)!) = 84 =FACT(9)/(FACT(6)*FACT(3))
(vi) 8P3 = 8!/(8-3)!) = 336 =FACT(8)/FACT(5)
(vii) 5P2 = 5!/(5-2)!) = 20 =FACT(5)/FACT(3)
(viii) 7C7 = 7!/(7!*(7-7)!) = 1 =FACT(7)/(FACT(7)*FACT(0))
(ix) 7P4 = 7!/(7-4)!) = 840 =FACT(7)/FACT(3)

Likely scenarios

(i) Number of ways of arranging 6 cars on a showroom floor.


(ii) Number of ways of arranging the seating plan of 8 persons,
consisting of 3 males and 5 females.
(iii) Number of sequences of visiting 9 stores, consisting of 4 clothing stores,
2 home décor stores and 3 coffee shops.
(iv) Selecting all combinations of 4 holiday destinations from a possible 7 destinations.
Similar for (v) and (viii)
(vi) Selecting a committee of 3 persons from 8 candidates, where the first person
selected is the chairman, the second is the secretary and the third is a member.
Similar for (vii) and (ix)
Exercise 4.16

Assume each advertisement contains a different combination of 7 out of 12 products.

12C7 = 12!/(7!*(12-7)!) = 792 different combinations

=FACT(12)/(FACT(7)*FACT(5)) (Using Excel )


Exercise 4.17

A different permutation of 3 soup brands on 5 shelves is required.

5P3 = 5!/(5-3)! = 60 distinct ordering of 3 soup brands on 5 shelves.

=FACT(5)/FACT(2) (Using Excel )


Exercise 4.18

(a) 9C4 = 9!/(4!*(9-4)!) = 126 separate portfolios of 4 equities.


=FACT(9)/(FACT(4)*FACT(5)) (Using Excel )

(b) P(3,5,7,8) = 1/126 = 0.007937 0,794% chance of getting this combination.


Exercise 4.19

No. of permutations of 5 screws = 5! 5.4.3.2.1 = 120

Thus the probability of replacing them in exactly the same order =


1/120 = 0.00833 (0,833% chance)
Exercise 4.20

(a) 10C3 = 10!/(3!*(10-3)!) = 120 different selections


of 3 tourist attractions from 10 options.
=FACT(10)/(FACT(3)*FACT(7)) (Using Excel )

(b) P(a given combination of 3 out of 10) = 1/120 = 0.0083 (0,833% chance)
Exercise 4.21

(a) 4 C2 x 7 C4 = (4!/(2!*2!))*(7!/(4!*3!)) = 210 different committees

=FACT(4)/(FACT(2)*FACT(2))*FACT(7)/(FACT(4)*FACT(3))
(Using Excel )

(b) (4C2 x 7C4) x 2 = (4!/(2!*2!))*(7!/(4!*3!)) x 2 = 420 different committees

=2*FACT(4)/(FACT(2)*FACT(2))*FACT(7)/(FACT(4)*FACT(3))
(Using Excel )
Exercise 4.22 Project Scoping Study Bayes Theorem

APPROACH 1 Using a PROBABILITY TREE

T On time S Scope change


L Late NS No scope change

Marginal probabilities Conditional probabilities Joint probabilities

P(S/T) = 0.4 P(S and T) = 0.28


P(T) = 0.7
On time P(NS/T) = 0.6 P(NS and T) = 0.42

Late P(S/L) = 0.8 P(S and L) = 0.24


P(L) 0.3
P(NS/L) = 0.2 P(NS and L) = 0.06

Bayes Application Using the Joint Probabilities from the Probability Tree

Given P(T) = 0.7 Prior Probability


Find P(T/S) = P(T and S)/P(S) = Posterior Probability

(i) P(S) = P(S and T) + P(S and L) = 0.52


(ii) P(T and S) = 0.28

Then P(T/S) = P(T and S)/P(S) = 0.5385

There is a 53.85% chance that a 'scope-changed' project will be completed on time.

---ooOoo---

APPROACH 2 Using TABLE FORMAT (Applying Marginals and Conditional Probabilities)

Additional Information
S NS
T 0.28 0.42 0.7 P(T/S) = P(T and S)/P(S)
L 0.24 0.06 0.3 =0.28/(0.28+0.24)
0.52 0.48 1 0.5385

---ooOoo---
Exercise 4.23 Married Couples Sporting Habits Study Bayes Theorem

APPROACH 1 Using a PROBABILITY TREE

HS Husband plays sport WS Wife plays sport


HN Husband does not play sport WNS Wife does not play sport

Marginal probabilites Conditional probabilities Joint probabilities

P(WS/HS) = 0.4 P(HS and WS) = 0.24


P(HS) = 0.6
Husband P(WNS/HS) = 0.6 P(HS and WNS) = 0.36
plays sport

P(HNS) 0.4 P(WS/HNS) = 0.3 P(HNS and WS) = 0.12


Husband does
not play sport P(WNS/HNS) = 0.7 P(HNS and WNS) = 0.28

Bayes Application Using the Joint Probabilities from the Probability Tree

Given P(HS) = 0.6 Prior Probability


Find P(HS/WS) = P(HS and WS)/P(WS) = Posterior Probability

(i) P(WS) = P(WS and HS) + P(WS and HNS) = 0.36


(ii) P(HS and WS) = 0.24

Then P(HS/WS) = P(HS and WS)/P(WS) = 0.6667

There is a 66,67% chance that a husband plays sport if the wife also plays sport.

---ooOoo---

APPROACH 2 Using TABLE FORMAT (Applying Marginals and Conditional Probabilities)

Additional Information
WS WNS
HS 0.24 0.36 0.6 P(HS/WS) = P(HS and WS)/P(WS)
HNS 0.12 0.28 0.4 =0.24/(0.24+0.12)
0.36 0.64 1 0.6667

---ooOoo---
Exercise 4.24 Airline Departure Times Study Bayes Theorem

APPROACH 1 Using a PROBABILITY TREE

A Airline A T Leaves on Time


B Airline B L Leaves Late

Marginal probabilities Conditional probabilities Joint probabilities

P(T/A) = 0.8 P(A and T) = 0.48


P(A) = 0.6
A P(L/A) = 0.2 P(A and L) = 0.12

B P(T/B) = 0.65 P(B and T) = 0.26


P(B) = 0.4
P(L/B) = 0.35 P(B and L) = 0.14

Bayes Application Using the Joint Probabilities from the Probability Tree

Given P(A) = 0.6 Prior Probability


Find P(A/T) = P(A and T)/P(T) = Posterior Probability

(i) P(T) = P(A and T) + P(B and T) = 0.74


(ii) P(A and T) = 0.48

Then P(A/T) = P(A and T)/P(T) = 0.6486

There is a 64.86% chance that it is Airline A, if the aircraft that has just left, left on time.

---ooOoo---

APPROACH 2 Using TABLE FORMAT (Applying Marginals and Conditional Probabilities)

Additional Information
T L P(A/T) = P(A and T)/P(T)
A 0.48 0.12 0.6 =0.48/(0.48+0.26)
B 0.26 0.14 0.4 0.6486
0.74 0.26 1

---ooOoo---
Exercise 4.25 New Business Venture Study Bayes Theorem

APPROACH 1 Using a PROBABILITY TREE

G NBV started by Graduate S Successful


NG NBV started by non-Graduate F Failure

Marginal probabilities Conditional probabilities Joint probabilities

P(S/G) = 0.8 P(G and S) = 0.48


P(G) = 0.6
Graduate P(F/G) = 0.2 P(G and F) = 0.12

Non
graduate P(S/NG) = 0.65 P(NG and S) = 0.26
P(NG) = 0.4
P(F/NG) = 0.35 P(NG and F) = 0.14

Bayes Application Using the Joint Probabilities from the Probability Tree

Given P(G) = 0.6 Prior Probability


Find P(G/F) = P(G and F)/P(F) = Posterior Probability

(i) P(F) = P(G and F) + P(NG and F) = 0.26


(ii) P(G and F) = 0.12

Then P(G/F) = P(G and F)/P(F) = 0.4615

There is a 46.15% chance that a NBV was started by a Graduate given that it has failed.

---ooOoo---

APPROACH 2 Using TABLE FORMAT (Applying Marginals and Conditional Probabilities)

Additional Information
S F
G 0.48 0.12 0.6 P(G/F) = P(G and F)/P(F)
NG 0.26 0.14 0.4 =0.12/(0.12+0.14)
0.74 0.26 1 0.4615

---ooOoo---
Exercise 4.26 On-line Airline Tickets Purchase Study Bayes Theorem

APPROACH 1 Using a PROBABILITY TREE

E e-Ticket purchase B Business Traveller


NE non e-Ticket purchase NB non-Business Traveller

Marginal probabilities Conditional probabilities Joint probabilities

P(B/E) = 0.8 P(E and B) = 0.32


P(E) = 0.4
e-Ticket P(NB/E) = 0.2 P(E and NB) = 0.08

non-
e-Ticket P(B/NE) = 0.45 P(NE and B) = 0.27
P(NE) = 0.6
P(NB/NE) = 0.55 P(NB and NB) = 0.33

1
Bayes Application Using the Joint Probabilities from the Probability Tree

Given P(E) = 0.4 Prior Probability


Find P(E/B) = P(E and B)/P(B) = Posterior Probability

(i) P(B) = P(E and B) + P(NE and B) = 0.59


(ii) P(E and B) = 0.32

Then P(E/B) = P(E and B)/P(B) = 0.5424

There is a 54.24% chance that an e-ticket was bought given that it was bought by a business traveller.

---ooOoo---

APPROACH 2 Using TABLE FORMAT (Applying Marginals and Conditional Probabilities)

Additional Information
B NB
E 0.32 0.08 0.4 P(E/B) = P(E and B)/P(B)
NE 0.27 0.33 0.6 =0.32/(0.32+0.27)
0.59 0.41 1 0.5424

---ooOoo---
CHAPTER 5

PROBABILITY DISTRIBUTIONS

Exercise 5.1 Binomial probability distribution


Poisson probability distribution

Exercise 5.2 (a) continuous (e.g. 35.142 gm)


(b) discrete (e.g. 132 employees)
(c) discrete (e.g. 7046 households)
(d) continuous (e.g. 514.68 km)
Exercise 5.3

3 (7-3)
(a) (i) n = 7; p = 0,2; x = 3 P(x = 3) = 7C3 (0.2) (0.8) =
0.1147 11.47%

4 (10-4)
(a) (ii) n = 10; p = 0,2; x = 4 P(x = 4) = 10C4 (0.2) (0.8) =
0.0881 8.81%

(a) (iii) n = 12; p = 0,3; x ≤ 4 P(x = 0) + P(x = 1) + P(x = 2) + P(x = 3) + P(x = 4)


0.01384+0.07118+0.16779+0.2397+0.23114 = 0.7237 72.37%

(a) (iv) n = 10; p = 0,05; x = 2 or 3 P(x = 2) + P(x = 3)


0.07464+0.01048 = 0.0851 8.51%

(a) (v) n = 8; p = 0,25; x ≥ 3 1 - (P(x = 0) + P(x = 1) + P(x = 2))


1 - (0.10011+0.26697+0.31146) = 0.3215 32.15%

(b) (i) =BINOMDIST(3,7,0.2,0) 0.1147


(b) (ii) =BINOMDIST(4,10,0.2,0) 0.0881
(b) (iii) =BINOMDIST(4,12,0.3,1) 0.7237
(b) (iv) =BINOMDIST(2,10,0.05,0)+BINOMDIST(3,10,0.05,0) 0.0851
(b) (v) =1-BINOMDIST(2,8,0.25,1) 0.3215
Exercise 5.4

(a) Binomial distribution


There are only two possible outcomes (in-stock; out-of-stock)
This outcome is observed 6 times (n = 6 stores)
The probability of observing the "out-of-stock" outcome, p = 0.20, is constant.
The stores (trials) are independent of each other

(b) P(x = 1) = 6C1 (0.2)1 (0.8)(6-1) = 0.3932 39.32% =BINOMDIST(1,6,0.2,0)

(c) P(x ≤ 2) = P(x = 0) + P(x = 1) + P(x = 2)


0.2621 + 0.3932 + 0.2458 = 0.9011 90.11% =BINOMDIST(2,6,0.2,1)

(d) P(x = 0) = 6C0 (0.2)0 (0.8)(6-0) = 0.2621 26.21% =BINOMDIST(0,6,0.2,0)

(e) Mean (binomial) = 6 x 0.2 = 1.2


On average, 1,2 stores out the 6 stores surveyed can be expected to be out of stock in a week.
Exercise 5.5

(a) Binomial distribution


n = 12 p = 0.15
P(x = 0) = 12C0 (0.15)0 (0.85)(12-0) = 0.1422 =BINOMDIST(0,12,0.15,0)

(b) n = 15 p = 0.15
P(x < 3) = P(x = 0) + P(x = 1) + P(x = 2) =
= 0.0874 + 0.2312 + 0.2857 = 0.6042 =BINOMDIST(2,15,0.15,1)
Exercise 5.6

(a) Binomial distribution


n = 10 p = 0.30 (probability of preferring the deluxe model)
P(x = 3) = 10C3 (0.30)3 (0.70)(10-3) = 0.2668 =BINOMDIST(3,10,0.3,0)

(b) n = 10 p = 0.70 (probability of preferring the standard model)


P(x > 2) = 1 - P(x ≤ 2) = 1 - (P(x = 0) + P(x = 1) + P(x = 2)) =
= 1 - (0.0000059 + 0.000138 + 0.00145) = 0.9984 =1-BINOMDIST(2,10,0.7,1)
Exercise 5.7

(a) Binomial distribution


n=8 p = 0.05 (probability of a defective Tata truck)
P(x = 1) = 8C1 (0.05)1 (0.95)(8-1) = 0.2793 =BINOMDIST(1,8,0.05,0)

(b) n=8 p = 0.05 (probability of a defective Tata truck)


P(x ≤ 2) = P(x = 0) + P(x = 1) + P(x = 2) =
= 0.66342 + 0.27934 + 0.05146 = 0.9942 =BINOMDIST(2,8,0.05,1)

(c) n=8 p = 0.05 (probability of a defective Tata truck)


P(x = 0) = 8C0 (0.05) (0.95)(8-0) = 0.6634 =BINOMDIST(0,8,0.05,0)
0

(d) Mean(binomial) = 64 x 0.05 = 3.2


Based on 64 sales, the dealer can expect 3.2 trucks to be returned for assemby defective repairs
Exercise 5.8

(a) Binomial distribution


n=6 p = 0.80 (probability of a UT out-performing the JSE Index)
P(x = 6) = 6C6 (0,80)6 (0,20)(6-6) = 0.2621 =BINOMDIST(6,6,0.8,0)

(b) n=6 p = 0.80 (probability of a UT out-performing the JSE Index)


P(x = (2 U 3)) = P(x = 2) + P(x = 3) =
= 0.01536 + 0.08192 = 0.0973
=BINOMDIST(2,6,0.8,0)+BINOMDIST(3,6,0.8,0)

(c) n=8 p = 0.20 (probability of a UT under-performing the JSE Index)


P(x ≤ 2) = P(x = 0) + P(x = 1) + P(x = 2) =
= 0.26214 + 0.39322 + 0.24576 = 0.9011 =BINOMDIST(2,6,0.2,1)
Exercise 5.9

(a) Binomial distribution


n = 12 p = 0.20 (probability of a person participating in a focus group)
P(x = 2) = 12C2 (0.20)2 (0.80)(12-2) = 0.2835
=BINOMDIST(2,12,0.2,0)

(b) n = 12 p = 0.20 (probability of a person participating in a focus group)


P(x = 5) = 12C5 (0.20) (0.80)(12-5) = 0.5320
5

=BINOMDIST(5,12,0.2,0)

(c) n = 12 p = 0.20 (probability of a person participating in a focus group)


P(x ≥ 6) = 1 - P(x ≤ 5) = 1 - (P(x=0)+P(x=1)+P(x=2)+P(x=3)+P(x=4)+P(x=5)) =
= 1 - (0.0687 + 0.2062 + 0.2835 + 0.2362 + 0.1329 + 0.0532) = 0.0194
=1 - BINOMDIST(5,12,0.2,1)
Exercise 5.10

(a) (i) Binomial distribution


n = 10 p = 0.10 (probability of a general population person being a heavy reader )
P(x < 2) = P(x=0)+P(x=1) = 10C0(0.10)0(0.90)(10-0) + 10C1(0.10)1(0.90)(10-1)
= 0.34868 + 0.38742 = 0.7361

P(x < 2) = =BINOMDIST(1,10,0.1,1)

(a) (ii) n = 10 p = 0.35 (probability of a pensioner person being a heavy reader )


P(x < 2) = P(x=0)+P(x=1) = 10C0(0.35)0(0.65)(10-0) + 10C1(0.35)1(0.65)(10-1)
= 0.01346 + 0.07249 = 0.0860

P(x < 2) = =BINOMDIST(1,10,0.35,1)

(b) n = 280 p = 0.65 (probability of a pensioner person not being a heavy reader )
Expected number (mean) of "non-heavy" readers = 280 x 0.65 = 182
Exercise 5.11

(a) Poisson distribution


P(x = 5 / a = 3) = e-3 35 / 5! = 0.1008 10.08% =POISSON(5,3,0)

(b) P(x ≥ 4 / a = 3) = 1 - P(x ≤ 3) = 1 - (P(x=0)+P(x=1)+P(x=2)+P(x=3))


-3 0 -3 1 -3 2 -3 3
= 1 - (e 3 / 0! +e 3 / 1! + e 3 / 2! + e 3 / 3!)
= 1 - (0.04979 + 0.14936 + 0.22404 + 0.22404)
= 1 - 0.64723 = 0.3528 35.28% =1-POISSON(3,3,1)

(c) P(x = 0 / a = 3) = e-3 30 / 0! = 0.0498 4.98% =POISSON(0,3,0)


Exercise 5.12

(a) Poisson distribution


P(x ≤ 2 / a = 4) = P(x=0)+P(x=1)+P(x=2)
-4 0 -4 1 -4 2
= e 4 / 0! +e 4 / 1! + e 4 / 2!
= 0,01832 + 0,07326 + 0,14653 = 0.2381 23.81% =POISSON(2,4,1)

(b) P(x ≥ 4 / a = 4) = 1 - P(x ≤ 3) = 1 - (P(x=0)+P(x=1)+P(x=2)+P(x=3))


= 1 - (0,2381 (from (i) above) + e-3 43 / 4!)
= 1 - (0,2381 + 0,19537)
= 1 - 0,43347 = 0.5665 56.65% =1-POISSON(3,4,1)
Exercise 5.13

(a) (i) Poisson distribution


P(x = 1 / a = 6) = e-6 61 / 1! = 0.0149 1.49% =POISSON(1,6,0)

(a) (ii) P(x ≤ 3 / a = 6) = P(x=0)+P(x=1)+P(x=2)+P(x=3)


-6 0 -6 1 -6 2 -6 3
= e 6 / 0! +e 6 / 1! + e 6 / 2! + e 6 / 3!
= 0,00248 + 0,01487 + 0,04462 + 0,08924 0.1512 15.12%
=POISSON(3,6,1)
(a) (iii) P(x ≥ 3 / a = 6) = 1 - P(x ≤ 2) = 1 - (P(x=0)+P(x=1)+P(x=2))
= 1 - (e-6 60 / 0! +e-6 61 / 1! + e-6 62/ 2!)
= 1 - (0,00248 + 0,01487 + 0,04462)
= 1 - 0,06197 = 0.938 93.80% =1-POISSON(2,6,1)

(b) Note: a = 3 since the mean orders must refer to a given half-day interval (i.e. 6/2 = 3)
-3 1
P(x = 1 / a = 3) = e 3 / 1! = 0.1494 14.94% =POISSON(1,3,0)

(c) Mean = 6 orders/day Std dev = √6 = 2.449 orders/day


Exercise 5.14

(a) Poisson distribution


P(x ≥ 3 / a = 1.8) = 1 - P(x ≤ 2) = 1 - (P(x = 0) + P(x = 1) + P(x = 2))
-1.8 0 -1.8 1 -1.8 2
= 1 - (e 1.8 / 0! + e 1.8 / 1! + e 1.8 / 2!)
= 1 - (0.1653 + 0.29754 + 0.26778)
= 1 - 0.7306 = 0.2694 26.94% =1-POISSON(2,1.8,1)

(b) P(x < 4 / a = 1.8) = P(x ≤ 3) = P(x = 0) + P(x = 1) + P(x = 2) + P(x = 3)


= e-1.8 1.80 / 0! + e-1.8 1.81 / 1! + e-1.8 1.82/ 2! + e-1.8 1.83/ 3!
= 0.1653 + 0.29754 + 0.26778 + 0.1607 = 0.8913 89.13%
=POISSON(3,1.8,1)
Exercise 5.15

(a) Poisson distribution


P(x ≤ 5 / a = 7) = P(x = 0) + P(x = 1) + P(x = 2) + P(x = 3) + P(x = 4) + P(x = 5)
= e-7 70 / 0! + e-7 71 / 1! + e-7 72/ 2! + e-7 73/ 3! + e-7 74 / 4! + e-7 75 / 5!
= 0.00091 + 0.00638 + 0.02234 + 0.05213 + 0.09123 +0.12772 =
0.3007 30.07% =POISSON(5,7,1)

(b) P(x = 6 or x = 9 / a = 7) = P(x=6) + P(x=9)


= e-7 76 / 6! + e-7 79 / 9! = 0.1490 + 0.1014 = 0.2504 25.04%
=POISSON(6,7,0)+POISSON(9,7,0)

(c) Note: a = 14 since the time interval is a given two-day period (i.e. 7 x 2 = 14)
P(x > 20 / a = 14) = 1 - P(x ≤ 20)
= 1 - (P(x = 0) + P(x = 1) + P(x = 2) + …… + P(x = 19) + P(x = 20))
Use Excel only.
=1 - POISSON(20,14,1) = = 1 - 0.9521 = 0.0479 4.79%
Exercise 5.16

Normal distribution Use the Standard Normal (z) table


(a)
(i) P(0 < z < 1.83) = 0.4664
(ii) P(z > -0.48) = 0.5 + P(0 < z < 0.48) = 0.5 + 0.1844 = 0.6844
(iii) P(-2.25 < z < 0) = P(0 < z < 2.25) = 0.4878
(iv) P(1.22 < z ) = 0.5 - P(0 < z < 1.22) = 0.5 - 0.3888 = 0.1112
(v) P(-2.08< z < 0.63) = P(0 < z < 2.08)+P(0 < z < 0.63) = 0,4812 + 0,2357 = 0.7169
(vi) P(z < -0.68) = 0.5 - P(0 < z < 0.68) = 0,5 - 0,2517 = 0.2483
(vii) P(0.33 < z < 1.5) = P(0 < z < 1.5) - P(0 < z < 0.33) = 0.4332 - 0.1293 = 0.3039

(b) Using Excel


(i) =NORMSDIST(1.83)-0.5 0.4664
(ii) =1-NORMSDIST(-0.48) 0.6844
(iii) =0.5-NORMSDIST(-2.25) 0.4878
(iv) =1-NORMSDIST(1.22) 0.1112
(v) =NORMSDIST(0.63)-NORMSDIST(-2.08) 0.7169
(vi) =NORMSDIST(-0.68) 0.2483
(vii) =NORMSDIST(1.5)-NORMSDIST(0.33) 0.3039
Exercise 5.17

Normal distribution Use the Standard Normal (z) table

(a) Look up area z-value


(i) P(z < ?) = 0.9147 0.9147 - 0.5 = 0.4147 1.37
(ii) P(z > ?) = 0.5319 0.5319 - 0.5 = 0.0319 -0.08
Since area to right of z is greater than 0.5, z must be negative.
(iii) P(0 < z < ?) = 0.4015 0.4015 1.29
(iv) P(? < z < 0) = 0.4803 0.4803 -2.06
(v) P(? < z ) = 0.0985 0.5 - 0.0985 = 0.4015 1.29
(vi) P(z < ?) = 0.2517 0.5 - 0.2517 = 0.2483 -0.67
Since area to left of z is less than 0.5, z must be negative.
(vii) P(? < z ) = 0.6331 0.6331 - 0.5 = 0.1331 -0.34
Since area to right of z is greater than 0.5, z must be negative.

(b) Using Excel


(i) =NORMSINV(0.9147) 1.3703
(ii) =NORMSINV(1-0.5319) -0.0800
(iii) =NORMSINV(0.5+0.4015) 1.2901
(iv) =NORMSINV(0.5-0.4803) -2.0600
(v) =NORMSINV(1-0.0985) 1.2901
(vi) =NORMSINV(0.2517) -0.6691
(vii) =NORMSINV(1-0.6331) -0.3401
Exercise 5.18

Use the Standard normal (z) table μ = 64 σ = 2.5


(a)
(i) P(x < 62) = P(z < (62-64)/2.5) = P(z < -0.08) = 0.5 - 0.2881 = 0.2119
(ii) P(x > 67.4) = P(z > (67.4-64)/2.5) = P(z > 1.36) = 0.5 - 0.4131 = 0.0869
(iii) P(59.6 < x < 62.8) = P((59.6-64)/2.5 < z < (62.8-64)/2.5) = P(-1.76 < z < -0.48) =
0.4608 - 0.1844 = 0.2764
(iv) P(x > ?) = 0.1026 Look up area = 0.3974 giving z ≈ 1.267
then x = 64+1.267*2.5 = 67.168 min
(v) P(x > ?) = 0.9772 Look up area = 0.9772 - 0.5 = 0.4772 giving z = -2.00
then x = 64-2.00*2.5 = 59 min
(vi) P(60.2 < x < ?) = 0.6652 Find P(60.2 < x < 64) = P(-1.52 < z < 0)
giving area = 0.4357.
Look up area = 0.6652 - 0.4357 = 0.2295 giving z ≈ 0.611
then x = 64+0.611*2.5 = 65.528 min

(b) Using Excel NORMDIST and NORMINV

(i) =NORMDIST(62,64,2.5,1) 0.2119


(ii) =1-NORMDIST(67.4,64,2.5,1) 0.0869
(iii) =NORMDIST(62.8,64,2.5,1)-NORMDIST(59.6,64,2.5,1) 0.2764
(iv) =NORMINV(1-0.1026,64,2.5) 67.167
(v) =NORMINV(1-0.9772,64,2.5) 59.002
(vi) =NORMDIST(60.2,64,2.5,1)+0.6652 0.729455
Apply area = 0.729455 to '=NORMINV(0.729455,64,2.5) 65.528
Exercise 5.19 Gym Attendance Duration

Normal distribution x ≡ N(80, 20)

(a) P(x > 120) = P(z > (120-80)/20) = P(z > 2) = 0.5 - 0.4772 = 0.0228 2.28%

(b) P(x < 60) = P(z (60-80)/20) = P(z < -1) = 0.5 - 0.3413 = 0.1587 15.87%

(c) P(x > k) = 0.60 Look up area = 0.60 - 0.50 = 0.10 giving z ≈ -0.253
Then x = 80 - 0.253*20 ≈ 74.94 min

Using Excel NORMDIST or NORMINV

(a) =1-NORMDIST(120,80,20,1) 0.0228

(b) =NORMDIST(60,80,20,1) 0.1587

(c) =NORMINV(1-0.6,80,20) 74.93 minutes


Exercise 5.20 Automatic Washing Machine Lifespan

Normal distribution x ≡ N(3.1; 1.1)

(a) P(x < 1) = P(z < (1-3.1)/1.1) = P(z < -1.9091) = 0.5 - 0.4719 = 0.0281 2.81%

(b) (i) P(x > 4) = P(z > (4-3.1)/1.1) = P(z > 0.819) = 0.5 - 0.2939 = 0.2061 20.61%

(b) (ii) P(x > 5.5) = P(z > (5.5-3.1)/1.1) = P(z > 2.182) = 0.5 - 0.4854 = 0.0146 1.46%

(c) P(x < k) = 0.05 Look up area = 0.50 - 0.05 = 0.45 giving z ≈ -1.645
Then x = 3.1 - 1.645*1.1 ≈ 1.29 years

Using Excel NORMDIST or NORMINV

(a) =NORMDIST(1,3.1,1.1,1) 0.0281

(b) (i) =1-NORMDIST(4,3.1,1.1,1) 0.2061

(b) (ii) =1-NORMDIST(5.5,3.1,1.1,1) 0.0146

(c) =NORMINV(0.05,3.1,1.1) 1.29 years


Exercise 5.21 Household Daily Water Usage

Normal distribution x ≡ N(220; 45)


(a)
(i) P(x > 300) = P(z > (300-220)/45) = P(z > 1.778) = 0.5 - 0.4625 = 0.0375 3.75%

(ii) P(x < 100) = P(z < (100-220)/45) = P(z < -2.667) = 0.5 - 0.4962 = 0.0038 0.38%

(iii) P(x < k) = 0.15 Look up area = 0.50 - 0.15 = 0.35 giving z ≈ -1.038 (approx)
Then x = 220 - 1.038*45 ≈ 173.29 litres

(iv) P(x > k) = 0,20 Look up area = 0.50 - 0.20 = 0.30 giving z ≈ 0.842 (approx)
Then x = 220 + 0.842*45 ≈ 257.89 litres

(b) Using Excel NORMDIST

(a) (i) =1-NORMDIST(300,220,45,1) 0.0377


(a) (ii) =NORMDIST(100,220,45,1) 0.0038

(c) Using Excel NORMINV

(a) (iii) =NORMINV(0.15,220,45) 173.36 litres


(a) (iv) =NORMINV(1-0.2,220,45) 257.87 litres
Exercise 5.22 Long-distance Truck Drivers' Reaction Times

(a) Normal distribution x ≡ N(1.4; 0.25) Variance = 0.0625 Std dev = 0.25

(i) P(x > 2) = P(z > (2-1.4)/0.25) = P(z > 2.4) = 0.5 - 0.4918 =
0.0082 0.82%
(ii) P(1.2 < x < 1.4) = P((1.2-1.4)/0.25 < z < (1.4-1.4)/0.25) = P(-0.80 < z < 0) =
0.2881 28.81%
(iii) P(x < 0.9) = P(z < (0.9-1.4)/0.25) = P(z < -2.00) = 0.5 - 0.4772 =
0.0228 2.28%
(iv) P(0.5 < x < 1.0) = P((0.5-1.4)/0.25 < z < (1.0-1.4)/0.25) = P(-3.6
0.49984
< z <--1.6)
0.4452
= =
0.0546 5.46%

(b) P(x > 1.8) = P(z > (1.8-1.4)/0.25) = P(z > 1.6) = 0.5 - 0.4452 =
0.0548 5.48%
No. of truck drivers = 120 * 0,0548 = 6,576 drivers 7 drivers (approx)

(c) P(x > 1.7) = P(z > (1.7-1.4)/0.25) = P(z > 1.2) = 0.5 - 0.3849 =
0.1151 11.51%
No. of truck drivers = 360 * 0.1151 = 41.436 drivers 42 drivers (approx)

(a) Using Excel NORMDIST

(i) =1-NORMDIST(2,1.4,0.25,1) 0.0082 0.82%


(ii) =0.5-NORMDIST(1.2,1.4,0.25,1) 0.2881 28.81%
(iii) =NORMDIST(0.9,1.4,0.25,1) 0.0228 2.28%
(iv) =NORMDIST(1,1.4,0.25,1)-NORMDIST(0.5,1.4,0.25,1) 0.0546 5.46%

(b) =1-NORMDIST(1.8,1.4,0.25,1) 0.0548 5.48%


truck drivers 6.576 7

(c) =1-NORMDIST(1.7,1.4,0.25,1) 0.115


truck drivers 41.425 42
Exercise 5.23 Hair dye Containers Mass

Normal distribution x ≡ N(18.2; 0.7) Variance = 0.49 Std dev = 0.7

(a) P(x < 18) = P(z < (18-18.2)/0.7) = P(z < -0.2857) = 0.5 - 0.1124 (approx) = 0.3876
or 38.76%

(b) P(x > k) = 0.15 Look up area = 0.50 - 0.15 = 0.35 giving z ≈ 1.038 (approx)
Then x = 18.2 + 1.038 * 0.7 ≈ 18.927 gm

Using Excel NORMDIST and NORMINV

(a) =NORMDIST(18,18.2,0.7,1) 0.3875 %


(b) =NORMINV(0.85,18.2,0.7) 18.926 gm
Exercise 5.24 Motor Vehicle Service Time

Normal distribution x ≡ N(70; 9) Variance = 81 Std dev = 9

(a) P(x < 60) = P(z < (60-70)/9) = P(z < -1.11) = 0.5 - 0.3665 = 0.1335 13.35%

(b) P(x > 90) = P(z > (90-70)/9) = P(z > 2.22) = 0.5 - 0.4868 = 0.0132 1.32%

(c) P(50 < x < 60) = P((50-70)/9 < z < (60-70)/9) = P(-2.22 < z < -1.11) =
0.4868 - 0.3665 = 0.1203 12.03%

(d) P(x > 80) = P(z > (80-70)/9) = P(z > 1.11) = 0.5 - 0.3665 = 0.1335 13.35%
No. of customers = 0.1335 * 80 = 10.68 customers

(e) P(z > (80-μ)/9) = 0.05 Look up area = 0.50 - 0.05 = 0.45 giving z ≈ 1.645
Then 1.645 = (80 - μ)/9 giving μ = 80 - 1.645 * 9 = 65.195 min

Using Excel NORMDIST

(a) =NORMDIST(60,70,9,1) 0.13326 13.33%

(b) =1-NORMDIST(90,70,9,1) 0.013134 1.31%

(c) =NORMDIST(60,70,9,1)-NORMDIST(50,70,9,1) 0.120126 12.01%

(d) =1-NORMDIST(80,70,9,1) 0.13326 13.33%

(e) This answer cannot be computed from the Excel function NORMDIST.
Exercise 5.25 Coffee Dispensing Machine - Cup Fill

Normal distribution x ≡ N(230; 10)

(a)
(i) P(x > 235) = P(z > (235-230)/10) = P(z > 0.5) = 0.5 - 0.1915 =
0.3085 30.85%
(ii) P(235 < x < 245) = P((235-230)/10 < z < (245-230)/10) = P(0.5 < z < 1.5) =
0.4332 - 0.1915 =
0.2417 24.17%
(iii) P(x < 220) = P(z < (220-230)/10) = P(z < -1.00) = 0.5 - 0.3413 =
0.1587 15.87%

(b) P(x > k) = 0.15 Look up area = 0.50 - 0.15 = 0.35 giving z ≈ 1.038 (approx)
Then x = 230 + 1.038 * 10 = 240.38 ml

(c) P(z < (220-μ)/10) = 0.10 Look up area = 0.50 - 0.10 = 0.40 giving z ≈ -1.282
Then -1.282 = (220 - μ)/10 giving μ = 220 + 1.282 * 10 = 232.82 ml

Using Excel NORMDIST and NORMINV

(a) (i) =1-NORMDIST(235,230,10,1) 0.3085 30.85%


(a) (ii) =NORMDIST(245,230,10,1)-NORMDIST(235,230,10,1) 0.2417 24.17%
(a) (iii) =NORMDIST(220,230,10,1) 0.1587 15.87%

(b) =NORMINV(1-.15,230,10) 240.36 ml


Exercise 5.26 Car Battery Lifespan

Normal distribution x ≡ N(28; 4)

(a) P(30 < x < 34) = P((30-28)/4 < z < (34-28)/4) = P(0.50 < z < 1.50) =
0.4332 - 0.1915 = 0.2417
or 24.17%

(b) P(x < 24) = P(z < (24-28)/4) = P(z < -1.00) = 0.5 - 0.3413 = 0.1587
or 15.87%

(c) P(x > k) = 0.60 Look up area = 0.60 - 0.50 = 0.10 giving z ≈ -0.253 (approx)
Then x = 28 - 0.253 * 4 ≈ 26.988 months

(d) P(z < (x-28)/4) = 0.05 Look up area = 0.50 - 0.05 = 0.45 giving z ≈ -1.645
Then -1.645 = (x - 28)/4 giving x = 28 - 1.645 * 4 = 21.42 months

Using Excel NORMDIST and NORMINV

(a) =NORMDIST(34,28,4,1) - NORMDIST(30,28,4,1) 0.2417 24.17%


(b) =NORMDIST(24,28,4,1) 0.1587 15.87%
(c) =NORMINV(0.4,28,4) 26.987 26.987 mths
(d) =NORMINV(.05,28,4) 21.421 21.421 mths
CHAPTER 6
SAMPLING AND SAMPLING DISTRIBUTIONS

6.1 To generalise sample findings to the target population.

6.2 Sampling methods; Concept of the sampling distribution.

6.3 subset.

6.4 representative.

6.5 Non-probability sampling: sample members are chosen using non-random criteria meaning
that some members of the target population are excluded from being included in the
sample.

Probability sampling: sample members are chosen using random selection processes so that
each target population member stands a chance of being included in the sample.

6.6 Random (probability) sampling methods. Every member of the target population has a
chance of being included in the sample. This is likely to result in a more representative
sample than if a non-probability sampling method was used.

6.7 Unrepresentativeness of the target population;

Not possible to measure sampling error.

6.8 random (chance)

6.9 equal

6.10 Simple random sampling

6.11 Systematic random sampling

6.13 Stratified random sampling

6.14 Cluster random sampling

6.15 It results in a smaller sampling error

6.16 sample statistic; population parameter

6.17 standard error

6.18 95.5%

6.19 Normal

6.20 n = 30 or larger

6.21 Central Limit Theorem.

6.22 Sampling error is the error made when estimating the true population parameter (e.g.
population mean) when using the sample statistic (e.g. sample mean).
CHAPTER 7

CONFIDENCE INTERVAL ESTIMATION

Exercise 7.1 To estimate a population parameter value by defining an Interval within which the true
population value is likely to fall at a stated level of confidence.

Exercise 7.2 Use the z-statistic since the population standard deviation, σ is known (Given σ = 8)

Standard error = 8/√64 = 1


95% Confidence level z(0.95) = 1.96 Use NORMSINV(0.975)
Margin of error z x SE 1.96

Lower 95% confidence limit 85 - 1.96(1) 83.04


Upper 95% confidence limit 85 + 1.96(1) 86.96

Exercise 7.3 t statistic

Exercise 7.4 Use the t -statistic since the population standard deviation, σ is unknown (Given s = 6)

Standard error (approx) = 6/√25 = 1.2


Degrees of freedom = 25 - 1 24
90% Confidence level t (0.10,24) = 1.711 Use TINV(0.10,24)
Margin of error t x SE 2.91 Note: TINV requires tail probability

Lower 90% confidence limit 54 - 1.711(1.2) 51.95


Upper 90% confidence limit 54 + 1.711(1.2) 56.05
Exercise 7.5

(a) Manually
x(bar) = 24.4 σ = 10.8 n = 144

std error = 10.8/√144 = 0.9


z crit (95%) = 1.96
95% Confidence level = 1.96 *0.9 = 1.764

Lower 95% confidence limit 24.4 - 1.764 = 22.636


Upper 95% confidence limit 24.4 + 1.764 = 26.164

Interpretation There is a 95% chance that the interval (22.636 to 26.164) covers
the actual average number of employees per SME in Gauteng.

(b) z-crit (0.95) =NORMSINV(0.975) = 1.96

(c) 95% Confidence level =CONFIDENCE(0.05,10.8,144) 1.764


Exercise 7.6

(a) Manually
x(bar) = 131.6 σ = 25 n = 87

std error = 2.68 25/√87 = 2.68


z crit (90%) = 1.645
90% Confidence level = 1.645*2.68 4.409

Lower 90% confidence limit 131.6 - 4.409 = 127.191


Upper 90% confidence limit 131.6 + 4.409 = 136.009

Interpretation
There is a 90% chance that the interval (127.191 to 136.009) covers the
actual average no. of palettes per order received by a sugar mill in Durban.

(b) z-crit (0.90) =NORMSINV(0.95) 1.645

(c) 90% Confidence level =CONFIDENCE(0.10,131.6,87) 4.409

(d) 90% confidence limits of total palettes shipped in a year

Lower 90% limit 127.191*720 = 91577.52 palettes


Upper 90% limit 136.009*720 = 97926.48 palettes

Interpretation
Total palettes shipped in a year, on average, is likely to be
between 91 577 and 97 926 palettes, with 90% confidence.
Exercise 7.7

(a) Manually
x(bar) = R356 σ = R44 n = 256

std error = 44/√256 = 2.75


z crit (95%) = 1.96
95% Confidence level = 1.96 * 2.75 = 5.39

Lower 95% confidence limit 356 - 5.39 = R350.61


Upper 95% confidence limit 356 + 5.39 = R361.39

Interpretation
There is a 95% chance that the interval (R350.61 to R361.39) covers the
actual average monthly car insurance premium of medium-sized cars.

(b) z crit (90%) = 1.645


90% Confidence level = 1.645*2.75 = 4.524

Lower 90% confidence limit 356 - 4,524 = R351.48


Upper 90% confidence limit 356 + 4,524 = R360.52

Interpretation
There is a 90% chance that the interval (R351.48 to R360.52) covers the
actual average monthly car insurance premium of medium-sized cars.

The 90% confidence interval limits are closer together than


the 95% confidence interval limits.
The lower confidence results in a more precise set of interval limits (narrower).

(c) z-crit (0.95) =NORMSINV(0.975) 1.96

(d) 95% Confidence level =CONFIDENCE(0.05,44,256) 5.3899

(e) 95% confidence limits of total monthly premium income

Lower 95% confidence limit 350.61*3000 R1 051 830


Upper 95% confidence limit 361.39*3000 R1 084 170

Interpretation
Total monthly premium income, on average, is likely to be between
R1 051 830 and R1 084 170 with 95% confidence.
Exercise 7.8

(a) Manually
x(bar) = 4.985 σ = 0.04 n = 50

std error = 0.04/√50 = 0.005657


z crit (99%) = 2.58
99% Confidence level = 2.58 * 0.005657 = 0.01460

Lower 99% confidence limit 4.985 - 0.0146 = 4.9704


Upper 99% confidence limit 4.985 + 0.0146 = 4.9996

Interpretation
There is a 99% chance that the interval (4.97 litres to 4.9996 litres)
covers the actual average volume of paint in all five-litre cans.

(b) Yes, the store owner has statistical evidence to confirm that the average
volume of paint in 5-litre cans is most likely to be below 5 litres.

(c) z-crit (0.99) =NORMSINV(0.995) 2.576

(d) 99% Confidence level =CONFIDENCE(0.01,0.04,50) 0.0146


Exercise 7.9

(a) Manually
x(bar) = 3.8 σ = 0.6 n = 24

std error = 0,122474 0.6/√24 = 0.122474


z crit (90%) = 1.645
90% Confidence level = 0.20147

Lower 90% confidence limit 3.8 - 0.20147 = 3.599


Upper 90% confidence limit 3.8 + 0.20147 = 4.001

Interpretation
There is a 90% chance that the interval (3.599 to 4.001) covers
the actual average inventory turnover rate of all convenience stores.

(b) z-crit (0.90) =NORMSINV(0.95) 1.645

(c) 90% Confidence level =CONFIDENCE(0.10,0.6,24) 0.20145


Exercise 7.10

(a) Manually
x(bar) = 166.2 s = 22.8 n = 21

std error (estimated) = 22.8/√21 = 4.975368


t crit (95%,20) 2.086
95% Confidence level = 10.3786

Lower 95% confidence limit 166.2 - 10.3786 155.82


Upper 95% confidence limit 166.2 + 10.3786 176.58

Interpretation There is a 95% chance that between 155.8 and 176.6 calls, on
average, will be received by the call centre daily.

(b) Manually
x(bar) = 166.2 s = 22.8 n = 21

std error (estimated) = 22.8/√21 = 4.975368


t crit (99%,20) 2.845
99% Confidence level = 14.1549

Lower 99% confidence limit 166.2 - 14.1549 152.05


Upper 99% confidence limit 166.2 + 14.1550 180.35

Interpretation There is a 99% chance that between 152.1 and 180.4 calls, on
average, will be received by the call centre daily.

(c) The 95% confidence interval is more precise (narrower), but less reliable than the
99% confidence interval which is less precise (wider), but more reliable.

(d) t-crit (0.95, 20) =TINV(0.05,20) 2.0860

(e) Over a 30 day period Lower 95% confidence limit 155.82*30 = 4674.60
Upper 95% confidence limit 176.58*30 = 5297.40

Interpretation There is a 95% chance that between 4675 and 5298 calls, on
average, will be received over a 30-day period.
Exercise 7.11

(a) Manually
x(bar) = 12.5 s = 3.4 n = 28

std error (estimated) = 3.4/√28 = 0.64254


t crit (90%, 27) 1.703
90% Confidence level = 1.0943

Lower 90% confidence limit 12.5 - 1.0943 = 11.406


Upper 90% confidence limit 12.5 + 1.0943 = 13.594

Interpretation There is a 90% chance that the actual mean dividend yield of
all JSE-listed companies lies between 11.41% and 13.59%.

(b) t-crit (0.90, 27) =TINV(0.10,27) 1.7033


Exercise 7.12

(a) Manually
x(bar) = 0.981 s = 0.052 n = 18

std error (estimated) = 0.052/√18 = 0.01226


t crit (99%, 17) = 2.8982
99% Confidence level = 0.0355

Lower 99% confidence limit 0.981 - 0.0355 = 0.9455


Upper 99% confidence limit 0.981 + 0.0355 = 1.0165

Interpretation There is a 99% chance that the average fill of 1-litre cartons
of milk lies between 0.9455 litres and 1.0165 litres.

Since the interval covers 1 litre, it can be concluded that


the cartons do contain one litre of milk on average.

(b) Manually
x(bar) = 0.981 s = 0.052 n = 18

std error (estimated) = 0.052/√18 = 0.01226


t crit (95%, 17) = 2.1098
95% Confidence level = 0.0259

Lower 95% confidence limit 0.981 - 0.0259 = 0.9551


Upper 95% confidence limit 0.981 + 0.0259 = 1.0069

Interpretation There is a 95% chance that the average fill of 1-litre cartons
of milk lies between 0.9551 litres and 1.0069 litres.

The 95% confidence interval is more precise (narrower), but less reliable than the
99% confidence interval which is less precise (wider), but more reliable.

(c) t-crit (0.99, 17) =TINV(0.01,17) 2.8982


t-crit (0.95, 17) =TINV(0.05,17) 2.1098
Exercise 7.13

(a) Manually
x(bar) = 1420 s = 160 n = 50

std error (estimated) = 160/√50 = 22.62742


t crit (90%, 49) = 1.676 (use df = 50 from Table 2, Appendix)
90% Confidence level = 37.9235

Lower 90% confidence limit 1420 - 37.9235 = 1382.08


Upper 90% confidence limit 1421 + 37.9235 = 1457.92

Interpretation There is a 90% chance that the average monthly wage of


union members lies between R1382.08 and R1457.92.

(b) Manually
x(bar) = 1420 s = 160 n = 50

std error (estimated) = 160/√50 = 22.62742


t crit (99%, 49) = 2.678 (use df = 50 from Table 2, Appendix)
99% Confidence level = 60.5962

Lower 99% confidence limit 1420 - 60.5962 = 1359.4


Upper 99% confidence limit 1420 + 60.5962 = 1480.6

Interpretation There is a 99% chance that the average monthly wage of


union members lies between R1359.40 and R1480.60.

(c) The 90% confidence interval is more precise (narrower), but less reliable than the
99% confidence interval which is less precise (wider), but more reliable.

(d) t-crit (0.90, 49) =TINV(0.10,49) 1.6766


t-crit (0.99, 49) =TINV(0.01,49) 2.6800
Exercise 7.14

x = 84 n = 200

p = 84/200 = 0.42
std error = √[(0.42)(0.58)/200] 0.0349
z crit (95%) = 1.96
95% Confidence level = 0.0684

Lower 95% confidence limit 0.42 - 0.0684 = 0.3516


Upper 95% confidence limit 0.42 + 0.0684 = 0.4884

Interpretation
There is a 95% chance that the percentage of manufacturing firms that
meet the employment equity charter lies between 35.2% and 48.8%.
Exercise 7.15

x = 68 n = 160

p = 68/160 = 0.425
std error = √[(0.425)(0.575)/160] = 0.03908
z crit (95%) = 1.96
95% Confidence level = 0.0766

Lower 95% confidence limit 0.425 - 0.0766 = 0.3484


Upper 95% confidence limit 0.425 + 0.0766 = 0.5016

Interpretation
There is a 95% chance that the percentage of cash-paying customers
lies between 34.8% and 50.2%.
Exercise 7.16

x = (365-78) = 287 n = 365

p = 287/365 = 0.786
std error = √[(0.786)*(0.214)/365] = 0.021467
z crit (90%) = 1.645
90% Confidence level = 0.03532

Lower 90% confidence limit 0.786 - 0.0353 = 0.7507


Upper 90% confidence limit 0.786 + 0.0353 = 0.8213

Interpretation
There is a 90% chance that the percentage of non-overdrawn cheque accounts
at the Tshwane branch lies between 75.1% and 82.1%.
Exercise 7.17

x = 120 n = 300

p = 120/300 = 0.4
std error = √[(0.4)*(0.6)/300] = 0.02828
z crit (90%) = 1.645
90% Confidence level = 0.04652

Lower 90% confidence limit 0.4 - 0.04652 = 0.3535


Upper 90% confidence limit 0.4 + 0.04652 = 0.4465

Interpretation
There is a 90% chance that the percentage of shoppers who frequent a shopping
mall primarily because of its store mix lies between 35.4% and 44.7%.
Exercise 7.18 File: X7.18 - cashier absenteeism.xlsx

(a) Descriptive Statistics - Absent Days (Using Data Analysis in Excel )

Absent Days
Mean 9.379
Standard Error 0.625
Median 9
Mode 7
Standard Deviation 3.364
Sample Variance 11.315
Kurtosis -0.919
Skewness -0.009
Range 12
Minimum 3
Maximum 15
Sum 272
Count 29
Confidence Level (95.0%) 1.280

The average number of days absent was 9.4 days.


68% of employees were absent between 6 and 12.7 days (within 1 std dev).
The data is symmetrical about the mean (skewness = -0.009).
Absenteeism per employee ranged between 3 (min) and 15 days (max) last year.

(b) Lower 95% Confidence Limit 9.379 - 1.28 = 8.10


Upper 95% Confidence Limit 9.379 + 1.28 = 10.66

Interpretation
There is a 95% chance that the average number of days absent last year by
all cashiers in a supermarket was between 8.1 days and 10.7 days.

(c) Since the 95% confidence interval covers 10 days (8.1 < μ < 10.66), it is
possible that the mean number of days absent per employee does exceed 10 days.
Thus the company's policy is not being strictly adhered to.

(d) t-crit (0.95, 28) =TINV(0.05,28) 2.0484


Given standard error = 0.6250
Confidence level (95%) = 0.625 * 2.0484 = 1.2803
Exercise 7.19 File: X7.19 - parcel masses.xlsx

(a) Descriptive Statistics - Parcel Weights (kg) (Using Data Analysis in Excel )

Parcel Masses (kg)


Mean 2.829
Standard Error 0.091
Median 2.78
Mode 3.22
Standard Deviation 0.5998
Sample Variance 0.360
Kurtosis -0.972
Skewness -0.108
Range 2.17
Minimum 1.75
Maximum 3.92
Sum 121.63
Count 43
Confidence Level (90.0%) 0.154

The average parcel weight was 2.83 kg.


68% of parcels weigh between 2.23 kg and 3.43 kg (within 1 std dev).
The data is symmetrical about the mean (skewness = -0.1076).
Parcel weights ranged between 1.75 kg (min) and 3.92 kg (max).

(b) Lower 90% Confidence Limit 2.829 - 0.154 = 2.675


Upper 90% Confidence Limit 2.829 + 0.154 = 2.983

Interpretation
There is a 90% chance that the average parcel weight is between 2.68 kg and 2.98 kg.

(c) Since the 90% confidence interval lies below 3 kg (2.675 < μ < 2.983), it is
highly likely that parcel weights do not exceed 3 kg.
Thus Post-Net is adhering to its policy.

(d) t-crit (0.90, 42) =TINV(0.10,42) 1.682


Given standard error = 0.0910
Confidence level (90%) = 0.091 * 1.682 = 0.154
Exercise 7.20 File: X7.20 - cost-to-income.xlsx

(a) Descriptive Statistics - Cost-to-Income Ratio (Using Data Analysis in Excel )

Cost-to-Income Ratio
Mean 71.240
Standard Error 1.997
Median 68
Mode 84
Standard Deviation 14.123
Sample Variance 199.451
Kurtosis -1.145
Skewness 0.139
Range 53
Minimum 44
Maximum 97
Sum 3562
Count 50
Confidence Level (95.0%) 4.014

The average cost-to-income ratio per company is 71.24%.


68% of companies have a cost-to-income ratio of between 57.1% and 85.4% (within 1 std dev).
The data is reasonably symmetrical about the mean (skewness = 0.139).
Cost-to-income ratios ranged between 44% and 97% for the sample of public companies.

(b) Lower 95% Confidence Limit 71.24 - 4.014 = 67.226


Upper 95% Confidence Limit 71.24 + 4.014 = 75.254

Interpretation
There is a 95% chance that the average cost-to-income ratio of public companies
is between 67.2% and 75.3%.

(c) t-crit (0.95, 49) =TINV(0.05,49) 2.0096


Given standard error = 1.9970
Confidence level (95%) = 1.997 * 2.0096 = 4.0132

(d) Find P(x > 75) using "μ" = 71.24 and "σ"= 14.123
Standardise x = 75 to t-stat t-stat = (75 - 71.24)/14.123 = 0.26623

Use Excel to find P(z > 0.26623)


=TDIST(0.26623,49,1) 0.39559

Interpretation
39.6% of all public companies are likely to have a cost-to-income ratio
in excess of the 75% rule of thumb.
Sample size determination

Exercise 7.21 z 1.96 n = 96


e 10
σ 50

Exercise 22
(a) z 2.58 n = 666
e 0.1
σ 1

(b) z 2.58 n = 296


e 0.15
σ 1

(c) z 2.58 n = 166


e 0.2
σ 1

Exercise 23 z 1.645 n = 752


p 0.5
e 0.03
CHAPTER 8

HYPOTHESIS TESTS

SINGLE POPULATION (MEANS, PROPORTIONS AND VARIANCES)

Exercise 8.1 To test whether a claim / statement made about a population parameter
value is probably true or false, based on sample evidence.

Exercise 8.2 The “closeness” of the sample statistics to the claimed population parameter value.

Exercise 8.3 The Five Steps of Hypothesis Testing


Step 1: Define the statistical hypotheses (the null and alternative hypotheses).
Step 2: Determine the region of acceptance of the null hypothesis.
Step 3: Compute the sample test statistic.
Step 4: Compare the sample test statistic to the region of acceptance.
Step 5: Draw the statistical and management conclusions.

Exercise 8.4 Level of significance (α)


(and sample size when the population standard deviation is unknown)

Exercise 8.5 Reject H0 in favour of H1 at the 5% level of significance.


Exercise 8.6

(i) H0: µ ≤ 560 H1: µ > 560


x (bar) = 577 σ = 86 n = 120 α = 0.05

(a) One-sided upper tailed test


Use z test statistic since σ is known.

Area of Acceptance z ≤ 1.645 Read off 0.45 from z-table; or


=NORMSINV(0.95) [using Excel ]
(b) z-stat = (577-560)/(86/√(120) = 2.165

Since z-stat (2.165) > z-crit (1.645), there is sufficient sample evidence at the
5% level of significance to reject H0 in favour of H1. (i.e. Reject H0)
Conclude that the population mean value is significantly larger than 560.

(c) p -value = P(z > 2.165) = 0.5 - 0.4848 = 0.0152 From z-table; or
=1-NORMSDIST(2.165)

(ii) H0: π ≥ 0.72 H1: π < 0.72


x = 216 n = 330 α = 0.10 Derive p = 216/330 = 0,6545

(a) One-sided lower tailed test


Use z test statistic for proportions

Area of Acceptance z ≥ -1.28 Read off 0.4 from z-table; or


=NORMSINV(0.9) [using Excel ]

(b) z-stat = (0.6545 - 0.72)/√[(0.72)(1-0.72)/330] = -2.65

Since z-stat (-2.65) < z-crit (-1.28), there is sufficient sample evidence at the
10% level of significance to reject H0 in favour of H1. (i.e. Reject H0)
Conclude that the true population proportion is significantly less than 0.72.

(c) p -value = P(z < -2.65) = 0.5 - 0.496 = 0.0040 From z-table; or
=NORMSDIST(-2.65)

(iii) H0: µ = 8.2 H1: µ ≠ 8.2


x (bar) = 9.6 s = 2.9 n = 30 α = 0.01

(a) Two-sided test


Use t test statistic since σ is unknown
(only given the sample standard deviation, s and n is small (n < 30))

Area of Acceptance -2.756 ≤ t ≤ +2.756 Read off T(0.005,29) from t-table;


or '=TINV(0.01,29) [using Excel ]

(b) t-stat = (9.6 - 8.2)/(2.9)/√(30) = 2.644

Since t-stat (2.644) falls within the area of acceptance, there is insufficient sample
evidence at the 1% level of significance to reject H0 in favour of H1. (i.e. Accept H0)
Conclude that the true mean value is equal to 8.2.

(c) p -value = 2 x P(t > 2.644) = 0.0131 =TDIST(2.644,29,2)


(iv) H0: µ ≥ 18 H1: µ < 18
x (bar) = 14.6 s = 3.4 n = 12 α = 0.01

(a) One-sided lower tailed test


Use t test statistic since σ is unknown
(only have the sample stardard deviation, s and n is small (n < 30))

Area of Acceptance t ≥ -2.718 Read off T(0.01,11) from t-table;


or '=TINV(0.02,11) [using Excel ]

(b) t-stat = (14.6 - 18)/(3.4)/√(12) = -3.464

Since t-stat (-3.464) < t-crit (-2.718), there is sufficient sample evidence at the
1% level of significance to reject H0 in favour of H1. (Reject H0)
Conclude that the true mean value is significantly below 18.

(c) p -value = P(t < -3.464) = 0.0026 '=TDIST(-(-3.464),11,1)

(v) H0: π = 0.32 H1: π ≠ 0.32


x = 68 n = 250 α = 0.05 Derive p = 68/250 = 0.272

(a) Two-sided test


Use z test statistic for proportions

Area of Acceptance -1.96 ≤ z ≤ +1.96 Read off z(0.975) from z-table; or


=NORMSINV(0.975) [Excel ]

(b) z-stat = (0.272 - 0.32)/√[(0.32)(1-0.32)/250] = -1.627

Since z-stat (-1.627) falls within the area of acceptance, there is insufficient sample
evidence at the 5% level of significance to reject H0 in favour of H1. (i.e. Accept H0)
Conclude that the true population proportion is equal to 0.32.

(c) p -value = 2 x P(z < -1.627) = 0.1037 '=2*NORMSDIST(-1.627)


Exercise 8.7

(a) H0: µ = 85 H1: µ ≠ 85 Two-sided test

(b) Use the z test statistic since σ is known (σ = 25 min (given))

(c) Use α = 0.05


Area of Acceptance -1.96 ≤ z ≤ 1.96 Read off 0.475 from z-table; or
=NORMSINV(0.975) [using Excel]

z-stat = (80.5 - 85)/(25/√(132) = -2.068

Statistical conclusion
Since z-stat (-2.068) lies below -z-crit (-1.96), there is sufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that the population mean value is significantly different from 85 minutes.
Visitors to the Knysna shopping mall do not spend 85 minutes on average in the mall.

(d) p -value = 2 x P(z < -2.068) = 0.0386 =NORMSDIST(-2.068) x 2

(e) Since p -value (0.0386) < α (0.05), there is strong sample evidence to
conclude that visitors to the Knysna shopping mall do not spend 85 minutes,
on average, in the mall.

By inspection of the sample mean, it appears that visitors to the Knysa


shopping mall spend significantly less than 85 minutes in the mall.
Exercise 8.8

(a) H0: µ ≥ 30 min H1: µ < 30 min One-sided lower tailed test

(b) Use the z test statistic since σ is known (σ = 10.5 min (given))

(c) Use α = 0.01


Area of Acceptance z ≥ -2.33 Read off 0.49 from z-table; or
=NORMSINV(0.01) [using Excel]

z-stat = (27.9 - 30)/(10.5/√(86)) = -1.855

Statistical conclusion
Since z-stat (-1.855) lies above z-crit (-2.33), there is insufficient sample evidence at
the 1% level of significance to reject H0 in favour of H1 (Do not reject H0).

Management conclusion
Conclude that the population mean value is at least 30 minutes.
The supermarket manager's belief is therefore valid.

(d) p -value = P(z < -1.855) = 0.0318 =NORMSDIST(-1.855)


Since p -value (0.0318) > α (0.01) the sample evidence does not refute H0.

The sample evidence is not strong enough to refute the belief that customers
spend 30 minutes or more, on average, doing their purchases at the supermarket.
Hence conclude that customers are likely to spend at least half-an-hour,
on average, in the supermarket doing their grocery shopping.
Exercise 8.9

(a) H0: µ ≤ 72 hours H1: µ > 72 hours One-sided upper tailed test

(b) Use the z test statistic since σ is known (σ = 18 hours (given))

Use α = 0.10
Area of Acceptance z ≤ 1.28 Read off 0.40 from z-table; or
=NORMSINV(0.90) [using Excel]

z-stat = (75.9 - 72)/(18/√(46)) = 1.470

Statistical conclusion
Since z-stat (1.47) > z-crit (1.28), there is sufficient sample evidence at
the 10% level of significance to reject H0 in favour of H1 (Reject H0).

Management conclusion
Conclude that the local importer's claim is valid.
Consignments are taking significantly longer than 72 hours to clear customs.

(c) p -value = P(z > 1.47) = 0.0708 =1-NORMSDIST(1.47)


Since p -value (0.0708) < α (0.10) this confirms support for H1.

There is moderate sample evidence (relative to α = 0.10) to conclude that


consignment clearance times, on average, are significantly longer than 72 hours.
Exercise 8.10

(a) Use the z test statistic since σ is known (σ = 14.7% (given))

(b) H0: µ ≤ 40% H1: µ > 40% One-sided upper tailed test

(c) Use α = 0.01


Area of Acceptance z ≤ 2.33 Read off 0.49 from z-table; or
=NORMSINV(0.99) [using Excel]

z-stat = (44.1 - 40)/(14.7/√(76)) = 2.431

Statistical conclusion
Since z-stat (2.431) > z-crit (2.33), there is sufficient sample evidence at
the 1% level of significance to reject H0 in favour of H1 (Reject H0).

Management conclusion
Conclude that the Department of Health's concern is justified.
The average markup is significantly greater than 40%.

(d) p -value = P(z > 2.431) = 0.0075 =1-NORMSDIST(2.431)

Since p -value (0.0075) << α (0.01) it confirms H1. There is overwhelming sample
evidence to conclude that the average % markup is significantly greater than 40%.
Exercise 8.11

(a) Use the t test statistic since σ is unknown (only s = 21 gms is given)

H0: µ = 700 gms H1: µ ≠ 700 gms Two-sided test

Use α = 0.05 with degrees of freedom = (n -1) = 63


Area of Acceptance -2.00 ≤ t ≤ 2.00 Read off t(0.025,63) from t-table; or
'=TINV(0.05,63) [using Excel]

t-stat = (695 - 700)/(21/√(64)) = -1.905

Statistical conclusion
Since t-stat (-1,905) lies within the acceptance area, there is insufficient sample
evidence at the 5% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

Management conclusion
Conclude that Ryeband Bakery is both legally compliant and not wasting ingredients.
The average weight of all white loaves produced by Ryeband Bakery is 700 gms.

(b) p -value = 2 x P(t < -1.905) = 0.0613 =TDIST(-(-1.905),63,2)

Since p -value (0.0613) > α (0.05) H0 cannot be rejected as the sample evidence
is weak in favour of H1.
Hence it can be concluded that, on average, the weight of white bread
loaves produced by the Ryeband Bakery is 700 gm.

(c) H0: µ ≥ 700 gms H1: µ < 700 gms One-sided lower tailed test

Use α 0.05 with degrees of freedom = (n -1) = 63


Area of Acceptance t ≥ -1.671 Read off t(0.05,63) from t-table; or
=TINV(0.10,63) [using Excel]

t-stat = (695 - 700)/(21/√(64)) = -1.905

Statistical conclusion
Since t-stat (-1.905) < t-crit (-1.671), there is sufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (i.e. Reject H0).

Management conclusion
Conclude that Ryeband Bakery is not legally compliant.The average weight of all
white bread loaves produced by Ryeband Bakery is significantly below 700 gms.
Exercise 8.12

(a) Use the t test statistic since σ is unknown (only s = R788 is given)

H0: µ ≥ 5500 H1: µ < 5500 One-sided lower tailed test

Use α = 0.10 with degrees of freedom = (n -1) = 17


Area of Acceptance t ≥ -1.33 Read off t(0.10,17) from t-table; or
=TINV(0.20,17) [using Excel]

t-stat = (5275 - 5500)/(788/√(18)) = -1.211

Statistical conclusion
Since t-stat (-1.211) lies within the acceptance area, there is insufficient sample
evidence at the 10% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

Management conclusion
Conclude that mean weekly sales of the new pudding flavour is not less than R5500.
The company should therefore not withdraw the product at this stage.

(b) p -value = P(t < -1.211) = 0.1212 =TDIST(-(-1.211),17,1)

Since p -value (0.1212) > α (0.10), H0 cannot be rejected as the sample evidence
is weak in favour of H1.
Hence it can be concluded that, on average, the weekly sales of the new pudding
flavour is not less than R5500 and the product should not be withdrawn.
Exercise 8.13

(a) Use the t test statistic since σ is unknown (only s = 3.6 is given)

H0: µ ≤ 80 kg H1: µ > 80 kg One-sided upper tailed test

Use α = 0.05 and degrees of freedom = (n -1) = 25


Area of Acceptance t ≤ 1.708 Read off t(0.05,25) from t-table; or
=TINV(0.10,25) [using Excel]

t-stat = (81.3 - 80)/(3.6/√(26)) = 1.841

Statistical conclusion
Since t-stat (1.841) > t-crit (1.708), it lies in the region of rejection.
Therefore there is sufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (Reject H0)

Management conclusion
Conclude that mean tensile strength of the consignment of wire is more than 80 kg.
Marathon Products should accept this consignment as it meets quality specs.

(b) p -value = P(t > 1.841) = 0.0388 =TDIST(1.841,25,1)

Since p -value (0.0388) < α (0.05) there is strong sample evidence to reject H0
in favour of H1.
Hence it can be concluded that, on average, the tensile strength of
the wire in the consignment exceeds 80 kg. Hence accept the consignment.
Exercise 8.14

(a) Use the t test statistic since σ is unknown (only s = 0.068 is given)

H0: µ ≥ 1 H1: µ < 1 One-sided lower tailed test

Use α = 0.05 and degrees of freedom = (n -1) = 19


Area of Acceptance t ≥ -1.729 Read off t(0.05,19) from t-table; or
=TINV(0.10,19) [using Excel]

std error = 0.068/√(20) = 0.0152


t-stat = (0.982 - 1)/(0.0152) = -1.1842

Statistical conclusion
Since t-stat (-1.1842) > t-crit (-1.729), it lies within the region of acceptance.
There is therefore insufficient sample evidence at the
5% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

Management conclusion
Conclude that the mean fill of one-litre milk containers is not less than 1 litre.
The Consumer Council's claim that containers are being underfilled is not valid.

(b) p -value = P(t < -1.1842) = 0.1255 =TDIST(-(-1.1842),19,1)

Since p -value (0.1255) > α (0.05), H0 cannot be rejected as there is no sample


evidence to support H1.
Hence it can be concluded that, on average, the mean fill of milk containers
is at least one 1 lite. Thus there is no statistical support for the claim.
Exercise 8.15

(a) Use the z test statistic for proportions

H0: π ≥ 0.30 H1: π < 0.30 One sided lower tailed test

Use α = 0.05
Area of Acceptance z ≥ -1.645 Read off z(0.45) from z-table; or
=NORMSINV(0.05) [using Excel ]

n= 400 x= 106
p= 106/400 = 0,265
std error = √(0.3*0.7)/400 = 0,0229
z-stat = (0.265 - 0.3)/(0.0229) = -1.5284

Statistical conclusion
Since z-stat (-1.5284) > z-crit (-1.645) there is insufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

Management conclusion
Conclude that 30% or more of listeners tune into the news broadcast of this station.
The company should advertise in this radio station's news timeslots.

(b) p -value = P(z < -1.5284) = 0.0632 '=NORMSDIST(-1.5284)

Since p-value (0.0633) > α (0.05), there is weak sample evidence to


reject H0 in favour of H1. (i.e. Accept H0)
The company can accept the radio station's claim as valid.

(c) p -value = P(z < -1.5284) = 0.0632 =NORMDIST(0.265,0.3,0.0229,1)


Exercise 8.16

(a) Use the z test statistic for proportions

H0: π ≤ 0.60 H1: π > 0.60

Use α = 0.05
Area of Acceptance z ≤ 1.645 Read off z(0.45) from z-table; or
=NORMSINV(0.95) [using Excel ]

n= 150 x= 54
p= (150-54)/150 = 0.64
std error = √(0.6*0.4)/150 = 0.04
z-stat = (0.64 - 0.6)/(0.04) = 1.000

Statistical conclusion
Since z-stat (1.00) < z-crit (1.645) there is insufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

Management conclusion
Conclude that not more than 60% of Cape Town motorists do not have vehicle insurance.
The motor vehicle advisor's claim is not valid.

(b) p -value = P(z > 1.000) = 0.1587 =1-NORMSDIST(1)

Since p -value (0.1587) > α (0.05), the sample evidence does not support H1.
Hence the insurance advisor's claim has no statistical validity.
Exercise 8.17

(a) Use the z test statistic for proportions

H0: π ≤ 0.15 H1: π > 0.15

Use α = 0.10
Area of Acceptance z ≤ 1.28 Read off z(0.40) from z-table; or
=NORMSINV(0.90) [using Excel]

n= 560 x= 96
p= 96/560 = 0.1714
std error = √(0,15*0,85)/560 = 0.0151
z-stat = (0,1714 - 0,15)/(0,0151) = 1.417

Statistical conclusion
Since z-stat (1.417) > z-crit (1.28) there is sufficient sample evidence at
the 10% level of significance to reject H0 in favour of H1 (i.e. Reject H0)

Management conclusion
Conclude that the churn rate in the telecommunications industry exceeds 15%.

(b) p -value = P(z > 1.417) = 0.0782 =1-NORMSDIST(1.417)


Since p -value (0.0782) > α (0.10), there is moderate sample evidence to
reject H0 in favour of H1.
The same management conclusion applies as in (a) above.

(c) p -value = P(z > 1.417) = 0,0782 =1-NORMDIST(0.1714,0.15,0.0151,1)


Exercise 8.18

(a) Use the z test statistic for proportions

H0: π ≥ 0.90 H1: π < 0.90

Use α = 0.01
Area of Acceptance z ≥ -2.33 Read off z(0.49) from z-table; or
=NORMSINV(0.01) [using Excel]

n= 300 x= 260
p= 260/300 = 0.8667
std error = √(0.9*0.1)/300 = 0.0173
z-stat = (0.8667 - 0.9)/(0.0173) = -1.9249

Statistical conclusion
Since z-stat (-1.9249) > z-crit (-2.33) there is insufficient sample evidence at
the 1% level of significance to reject H0 in favour of H1 (i.e. Do not reject H0)

Management conclusion
Conclude that the germination rate of the barley seed is at least 90%.
The cooperative can justify buying the barley seed from this merchant.
The cooperative can accept the seed merchant's claim.

(b) p -value = P(z < -1.9249) = 0.0271 =NORMSDIST(-1.9249)


Since p -value (0.0271) > α (0.01), there is weak sample evidence at
the 1% level of signficance to reject H0 in favour of H1. Hence do not reject H0.
The same management conclusion applies as in (a) above.

(c) p -value = P(z < -1.9249) = 0.0271 =NORMDIST(0.8667,0.9,0.0173,1)


Exercise 8.19 File: X8.19 - cost-to-income.xlsx

Use the t test statistic since σ is unknown


Cost-to-income (%)
H0: µ ≥ 75% H1: µ < 75% Mean 71.24
Standard Error 1.9973
Use α = 0.05 and degrees of freedom = (n -1) = 49 Median 68
Read off t(0.05,49) from t-table; or Mode 84
=TINV(0.10,49) [using Excel] Standard Deviation 14.1227
Sample Variance 199.451
Area of Acceptance t ≥ -1.676 Kurtosis -1.1450
Skewness 0.1391
std error = 1.9973 Range 53
Minimum 44
t-stat = (71.24 - 75)/(1.9973) = -1.8825 Maximum 97
Sum 3562
Count 50

Statistical conclusion
Since t-stat (-1.8825) < t-crit (-1.676) there is sufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (i.e. Reject H0)

Management conclusion
Conclude that the mean cost-to-income ratio of JSE companies is less than 75%.
JSE companies are therefore adhering to the rule of thumb.
Exercise 8.20 File: X8.20 - kitchenware.xlsx

(a) Normality assumption check

Sales values Count


≤ 75 5
76 - ≤ 100 6
101 - ≤ 125 8
126 - ≤ 150 11
151 - ≤ 175 9
176 - ≤ 200 6
201 - ≤ 225 4
> 225 1

The distribution appears to be normal.

(b) Use the t test statistic since σ is unknown Sales Transaction value
Mean 137.12
H0: µ ≥ R150 H1: µ < R150 Standard Error 6.657
Median 138.5
Use α = 0.05 and degrees of freedom = (n -1) = 49 Mode 184
Read off t(0.05,49) from t-table; or Standard Deviation 47.074
=TINV(0.10,49) [using Excel] Sample Variance 2215.944
Kurtosis -0.741
Area of Acceptance t ≥ -1.676 Skewness 0.022
Range 190
std error = 6.657 Minimum 52
Maximum 242
t-stat = (137.12 - 150)/(6.657) = -1.9348 Sum 6856
Count 50

Statistical conclusion
Since t-stat (-1.9348) < t-crit (-1.676) there is sufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (Reject H0).

Management conclusion
Conclude that the mean transaction value at the Claremont branch is signficantly
below R150.
The management are advised to close the Claremont branch since it is unprofitable.
Exercise 8.21 File: X8.21 - flight delays.xlsx

(a) Normality assumption check

delays Count
≤5 0
5 - ≤ 7.5 8
7.6 - ≤ 10 28
10.1 - ≤ 12.5 29
12.6 - ≤ 15 13
15.6 - ≤ 17.5 2
> 17.5 0

Histogram of Flight Delay Times


35

28 29
30

25
No. of Flights

20

15 13

10 8

5 2
0 0
0
5 7.5 10 12.5 15 17.5 More

Delay time intervals

The histogram distribution appears normal. The assumption is satisfied.

(b) Use the t test statistic since σ is unknown flight delays


Mean 10.324
H0: µ ≤ 10 min H1: µ > 10 min Standard Error 0.261
Median 10.3
Use α = 0.10 and degrees of freedom = (n -1) = 79 Mode 11
Read off t(0.10,79) from t-table; or Standard Deviation 2.333
=TINV(0.20,79) [using Excel] Sample Variance 5.444
Kurtosis -0.550
Area of Acceptance t ≥ 1.292 Skewness 0.084
Range 10.3
std error = 0.261 Minimum 5.3
Maximum 15.6
t-stat = (10.324 - 10)/(0.261) = 1.241 Sum 825.9
Count 80

Statistical conclusion
Since t-stat (1.241) < t-crit (1.292) there is insufficient sample evidence at
the 10% level of significance to reject H0 in favour of H1 (i.e. Accept H0).

Management conclusion
Conclude that flight delay times, on average, do not exceed 10 minutes.
ACASA management do not need to conduct an indepth investigation on flight delays.
Exercise 8.22 File: X8.22 - medical claims.xls

(a) Normality assumption check

Claims intervals Count


≤ 100 5
101 - ≤ 125 3
126 - ≤ 150 9
151 - ≤ 175 21
176 - ≤ 200 25
201 - ≤ 225 18
226 - ≤ 250 8
251 - ≤ 300 11

Histogram of Daily Claims Processed

30
25
25
21
20 18
No. of Claims

15
11
9
10 8
5
5 3
0
0
100 125 150 175 200 225 250 300 More

Claims processed intervals

The distribution appears to be reasonably normal. The assumption is satisfied

(b) Use the t test statistic since σ is unknown Medical Claims


Mean 190.39
H0: µ ≤ 180 claims H1: µ > 180 claims Standard Error 4.568
Median 190
Use α = 0.01 and degrees of freedom = (n -1) = 99 Mode 210
Read off t(0.01,99) from t-table; or Standard Deviation 45.680
=TINV(0.02,99) [using Excel] Sample Variance 2086.685
Kurtosis -0.092
Area of Acceptance t ≥ 2.365 Skewness 0.096
Range 199
std error = 4.568 Minimum 92
Maximum 291
t-stat = (190.39 - 180)/(4.568) = 2.275 Sum 19039
Count 100

Statistical conclusion
Since t-stat (2.275) < t-crit (2.365) there is insufficient sample evidence at
the 1% level of significance to reject H0 in favour of H1 (Thus accept H0).

Management conclusion
Conclude that the average number of claims processed daily does not exceed 180.
Thus the supervisor has no statistical grounds for requesting additional staff.
Exercise 8.23 File: X8.23 - newspaper readership.xlsx

Guardian Newspaper Claim

(a) One-Way Pivot Table (Frequency Table) and Bar Chart

Tabloid Data Total


Sun % 15.8%
Count 19
Guardian % 35.0% Tabloid %
Count 42 Sun 15.8
Mail % 25.8% Guardian 35
Count 31 Mail 25.8
Voice % 23.3% Voice 23.3
Count 28
Total % 100%
Total Count 120

Bar Chart of Tabloid Readership

40
35
35
30
25.8
23.3
% of Readers

25
20
15.8
15
10
5
0
Sun Guardian Mail Voice

Interpretation
Based on the sample fo 120 tabloid readers, the Guardian newspaper has the
largest share of 35%. The Sun has the lowest percentage of readers at 16%.

(b) Use the z test statistic for proportions

H0: π ≥ 0.40 H1: π < 0.40 Test claim that Guardian newspaper
has at least a 40% market share.
(c) Use α = 0.05
Area of Acceptance z ≥ -1.645 Read off z(0.45) from z-table; or
=NORMSINV(0.05) [using Excel]

n= 120 x= 42
p= 42/120 = 0.35
std error = √(0.4*0.6)/120 = 0.0447
z-stat = (0.35 - 0.4)/(0.0447) = -1.1186

Statistical conclusion
Since z-stat (-1.1186) > z-crit (-1.645) there is insufficient sample evidence at
the 5% level of significance to reject H0 in favour of H1 (Thus accept H0)
Management conclusion
Hence conclude that the Guardian's market share is at least 40%
The Guardian newspaper 's claim is justified statistically.
Exercise 8.24 File: X8.24 - citrus products.xlsx

(a) One-Way Pivot Table (Frequency Table) and Bar Chart

Awareness Data Total


High Count 34
% 20% Awareness %
Low Count 72 Low 42.35
% 42.35% Moderate 37.65
Moderate Count 64 High 20
% 37.65%
Total Count 170
Total % 100%

Bar Chart of Awareness Levels

50
42.35
37.65
% of respondents

40

30
20
20

10

0
Low Moderate High

Awareness Levels

Interpretation
80% of sampled consumers have a low or moderate awareness. Only 20% of
the sampled consumers indicated a high awarness of the nutritional benefits of citrus.

(b) (i) Use the z test statistic for proportions

H0: π ≤ 0.15 H1: π > 0.15

(b) (ii) Use α = 0.01


Area of Acceptance z ≤ 2.326 Read off z(0.49) from z-table; or
=NORMSINV(0.99) [using Excel]

n= 170 x= 34
p= 34/170 = 0.2
std error = √(0.15*0.85)/170 = 0.0274
z-stat = (0.20 - 0.15)/(0.0274) = 1.825

Statistical conclusion
Since z-stat (1.825) < z-crit (2.326) there is insufficient sample evidence at
the 1% level of significance to reject H0 in favour of H1 (i.e. Accept H0)

(b) (iii) Management conclusion


Since the level of high consumer awareness does not exceed 15%, it is recommended that
Fruitco should launch a national awareness campaign.
Exercise 8.25 File: X8.25 - aluminium scrap.xlsx

(a) Histogram of Daily % Scrap of Machine

Daily % Scrap Count


≤ 2.8 6
2.8 - ≤ 3.2 12
3.2 - ≤ 3.6 13
3.6 - ≤ 4.0 11
4.0 - ≤ 4.4 5
4.4 - ≤ 4.8 3
Total 50

Histogram of Machine Daily % Scrap

16

14 13
12
12 11

10
No. of days

8
6
6 5

4 3

Scrap % intervals

Interpretation
The assumption of normality is largely satisfied.
The histogram is only moderately skewed to the right.

(b) 95% Confidence Limits for Machine

Machine - Daily % Scrap


Mean 3.483
Standard Error 0.0723
Median 3.41
Mode 2.97
Standard Deviation 0.511
Sample Variance 0.261
Kurtosis -0.851
Skewness 0.331
Range 1.83
Minimum 2.67
Maximum 4.5
Sum 174.14
Count 50
Confidence Level(95.0%) 0.145

Lower 95% confidence limit = 3.483 -0.145 = 3.34


Upper 95% confidence limit = 3.483+0.145 = 3.63
There is a 95% chance that the average daily % scrap produced by the machine
is likely to lie between 3.34% and 3.63%.

(c) Let μ = population mean daily % scrap produced by the machine.


Use the t test statistic since σ is unknown (only s is given)

H0: μ ≥ 3.75% One sided lower tailed test


H1: μ < 3.75%

Region of Acceptance Use α = 0.05 with df = (50 - 1) = 49


t-crit = t(0.05)(49) = =TINV(0.1,49) -1.677

Decision rule Do not reject H0 in favour of H1 if -1.677 ≤ t-stat

t-stat std error = 0.0723 (see Table above)

t-stat = (3.483 - 3.75)/0.0723


-3.693

Statistical conclusion
Since t-stat (-3.693) lies well within the region of rejection, there is strong sample
evidence at the 5% significance level to reject H0 in favour of H1. (Thus reject H0).

Management conclusion
Conclude that the average daily % scrap produced by the machine is less than 3.75%.
The machine is not yet due for a full maintenance service.
Exercise 8.26 Hypothesis Test for a Single Population Variance, σ2
5m pipe length variability analysis
Problem Characteristics
Variable x = length of pipe (specification = 5m)
Data type numerical, ratio scaled, continuous.

Data n = 26 pipes
α= 0.05
s= 3.46 σ0 = 3

Step 1 H0 : σ2 ≤ 9 One sided Upper tailed test


H1 : σ2 > 9 Management Question in H0

Step 2 Region of Non-Rejection (Use Chi-Square)


2
Upper Χ -crit = 37.652 Do not reject Ho if Χ2-stat ≤ 37.652

2
Step 3 Sample test statistic X -stat Formula 8.4
2
X -stat = 33.254 Also p -value = 0.12
Use CHISQ.DIST.RT(x, df)
Steps 4, 5 Statistical Conclusion
Since X2-stat < upper X2-crit (i.e. lies in region of non-rejection of H0),
do not reject H0 at 5% level of significance.

Management Conclusion
The production manager can be 95% confident that the variation
in pipe lengths is within the limits of the product specification.
Exercise 8.27 Hypothesis Test for a Single Population Variance, σ2
Problem Characteristics
Variable x = unknown - but numeric
Data type numerical, ratio scaled, continuous

Data n = 20
α= 0.1
s2 = 49.3 σ20 = 30

Step 1 H0 : σ2 ≤ 30 One sided Upper tailed test


2
H1 : σ > 30

Step 2 Region of Acceptance (Use Chi-Square)


2
Upper Χ -crit = 27.204 2
Do not reject Ho if Χ -stat ≤ 27.204

2
Step 3 Sample test statistic X -stat Formula 8.4
X2-stat = 31.223 Also p -value = 0.0382
Use CHISQ.DIST.RT(x, df)
Steps 4, 5 Statistical Conclusion
Since X2-stat > X2-crit (i.e. lies in region of rejection of H0),
reject H0 at 10% level of significance.

Management Conclusion
We are 90% confident that the population variance is significantly greater
than the specified value of 30.
Exercise 8.28 Hypothesis Test for a Single Population Variance, σ
2

Insurance claim values variability analysis


Problem Characteristics
Variable x = claim values (in Rand)
Data type numerical, ratio scaled - implies means and standard deviations

Data n = 32 claims
α= 0.05
2
s= 84 σ 0= 5625

2
Step 1 H0 : σ = 5625 Two-tailed test
2
H1 : σ ≠ 5625 Management Question in H0

Step 2 Region of Non-Rejection (Use Chi-Square)


2 2
Upper Χ -crit = 48.232 Lower Χ -crit = 17.539
2
Do not reject Ho if 17.539 ≤ Χ -stat ≤ 48.232

2
Step 3 Sample test statisticX -stat Use Formula 8.4
2
X -stat = 38.886 Also p -value = 0.1561 Use CHISQ.DIST.RT(x, df)

Steps 4, 5 Statistical Conclusion


Since X2-stat < X2-crit (i.e. lies in region of non-rejection of H0),
do not reject H0 at 5% level of significance.

Management Conclusion
2
The insurance industry can be 95% confident that the variation in claim values is still σ = R5625.

---ooOoo---
Exercise 8.29 File: X8.29 - pain relief.xlsx

Problem Characteristics
Variable x = time to pain relief (in minutes)
Data type numerical, ratio scaled, continuous

Data n = 16 patients
α= 0.01
s= 0.9004 σ20 = 1.8

Step 1 H0 : σ2 ≥ 1.8 One sided lower tailed test


2
H1 : σ < 1.8 Management Question in H0

Step 2 Region of Non-Rejection (Use Chi-Square)


2
Lower Χ -crit = 5.229 2
Do not reject Ho if Χ -stat ≥ 5.299

2
Step 3 Sample test statistic X -stat Use Formula 8.4
2
X -stat = 6.756 Also p -value = 0.0359
Use CHISQ.DIST(x, df)
Steps 4, 5 Statistical Conclusion
Since X2-stat > lower X2-crit (i.e. lies in region of non-rejection of H0),
do not reject H0 at 1% level of significance.

Management Conclusion The pharmaceutical company can be 99% confident that


the variation in time to pain relief from the new headache pill is not significantly less than
2
the current headache pill (of σ = 1.8 min).
(i.e. it does not significantly reduce variation in time to pain relief).
CHAPTER 9

HYPOTHESIS TESTS

COMPARISON BETWEEN TWO POPULATIONS (MEANS, VARIANCES AND PROPORTIONS)

Exercise 9.1 When the population standard deviations of the two populations are unknown.

Exercise 9.2 Use Formula 9.1


z-stat = [(72 - 66) - 0]/√(202/40+102/50)
z-stat = 1.7321

Exercise 9.3 (a) t-crit (0.05,38) = 2.024 Refer to Appendix 1 Table 2


(b) t-crit (0.05,90) = 1.987 Refer to Appendix 1 Table 3

Exercise 9.4 When the two samples are not independent.

Exercise 9.5 H0: π1 ≥ π2


H1: π1 < π2 One sided lower tailed test.
Exercise 9.6

(a) Test for equality of variances

H0 : σ21 = σ22 Two tailed test


H1 : σ²1 ≠ σ²2

F-cri t = F(0.05/2, 23,18) = 2.50 (from Table 4(b), Appendix 1) =F.INV.RT(0.025,23,18)


Decision rule: Do not reject H0 if F-stat ≤ 2.50 2.515

F-stat = = 4.14²/3.32² = 1.555 Use rule of larger variance in numerator

Since F-stat (1.555) < F-crit (2.515), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Therefore, use the pooled-variances t -test to test for differences in means.

(b) Test for difference between two population means

Let population 1 = Manufacturers


Let population 2 = Retailers
Let μi = population mean earnings yield (%) per sector i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0 : μ1 = μ2 Two tailed test


H1 : μ1 ≠ μ2

(c) Region of Acceptance Use α = 0.05 with df = (19+24-2) = 41


t-crit = t(0.05)(41) = 2.021

Decision rule Do not reject H0 in favour of H1 if -2.021 ≤ t-stat ≤ 2.021

t-stat s² (pooled variance) = ((19-1)3,32²+(24-1)4,14²)/(19+24-2)


14.454

std error = √(14.454*(1/19+1/24))


1.16747

t-stat = ((8.45-10.22)-(0))/1.16747
-1.51609

Conclusion
Since t-stat (-1.51609) lies within the region of acceptance, there is
insufficient sample evidence at the 5% level of significance to
reject H0 in favour of H1. (i.e. Accept H0).
Conclude that there is no difference in the mean earnings yield (%) between
manufacturing companies and retail companies.
Exercise 9.7

(a) Let population 1 = DIY Consumers


Let population 2 = Non-DIY Consumers
Let μi = population mean age of each consumer group i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≥ μ2 One sided lower tailed test


H1: μ1 < μ2

(b) Region of Acceptance Use α = 0.10 with df = (29+34-2) = 61


t-crit = t(0.10)(61) = -1.296

Decision rule Do not reject H0 in favour of H1 if -1.296 ≤ t-stat

t-stat s² (pooled variance) = ((29-1)15.9²+(34-1)16.2²)/(29+34-2)


258.0197

std error = √(258.0197*(1/29+1/34))


4.060301

t-stat = ((41.8-47.4)-(0))/4.0603
-1.379

(c) Statistical conclusion


Since t-stat (-1.379) lies outside (below) the region of acceptance,
there is sufficient sample evidence at the 10% level of significance
to reject H0 in favour of H1. (i.e. Reject H0).
Management conclusion
Conclude that the mean age of DIY consumers is significantly lower than
the mean age of non-DIY consumers at the 10% level of significance.

(d) Region of Acceptance (new) Use α = 0.05 with df = (29+34-2) = 61


t-crit = t(0.05)(61) = -1.671
Decision rule Do not reject H0 in favour of H1 if -1.671 ≤ t-stat

Statistical conclusion
t-stat (-1.379) now lies within the (new) region of acceptance.
Thus there is insufficient sample evidence at the 5% level of significance
to reject H0 in favour of H1. (i.e. Accept H0).
Management conclusion
Conclude that there is no significant difference in the mean ages between
DIY consumers and non-DIY consumers.
Exercise 9.8

(a) Let population 1 = Bus commuters


Let population 2 = Train commuters
Let μi = population mean commuting time for each transport mode i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≤ μ2 One sided upper tailed test


H1: μ1 > μ2

(b) Test for equality of variances

2 2
H0: σ 1 =σ 2 Two tailed test
H1: σ²1 ≠ σ²2

F-cri t = F(0.05/2, 21,35) = 2.10 (from Table 4(b), Appendix 1) =F.INV.RT(0.025,21,35)


Decision rule: Do not reject H0 if F-stat ≤ 2.10 2.105

F-stat = = 7.8²/4.6² = 2.875 Use rule of larger variance in numerator

Since F-stat (2.875) > F-crit (2.10), reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are different.


Therefore, use the unequal-variances t -test to test for differences in means.

(c) Region of Acceptance df = 30 df (num) = 11.14938


(Use df formula 9.9) df (denom) =0.374049
Use α = 0.01 with df = 30
t-crit = t(0.01)(30) = 2.75

Decision rule Do not reject H0 in favour of H1 if t-stat ≤ 2.75

t-stat std error = √(7.82/22 + 4.62/36)


1.680545

t-stat = ((35.3-31.8)-(0))/1.6805
2.083

Statistical conclusion
Since t-stat (2,083) lies within the region of acceptance,
there is insufficient sample evidence at the 1% level of significance
to reject H0 in favour of H1. (i.e. Accept H0).

(d) Management conclusion


Conclude that there is no difference in the mean commuting times between
bus and train commuters.
Recommendation
Since there is no difference in mean commuting times, either mode of transport can be
prioritised for upgrading (or both can be upgraded simultaneously).
Exercise 9.9

Let population 1 = Mastercard users


Let population 2 = Visa Card users
Let μi = population mean month-end credit card balance (in Rands) for each card type i .

Use the z test statistic since σi's are known

H0: μ1 = μ2 Two sided test


H1: μ1 ≠ μ2

Region of Acceptance Use α = 0.05


z-crit = z(0.05) = 1.96

Decision rule Do not reject H0 in favour of H1 if -1.96 ≤ z-stat ≤ 1.96

z-stat std error = √(294²/45+336²/66)


60.26065

z-stat = ((922-828)-(0))/60.26065
1.5599

Statistical conclusion
Since z-stat (1.5599) lies within the region of acceptance, there is insufficient sample
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that there is no difference in the mean month-end credit card balances
between Mastercard holders and Visa Card holders.
Exercise 9.10

Let population 1 = non-attendees of job enrichment workshops


Let population 2 = attendees of job enrichment workshops
Let μi = population mean job satisfaction rating for each employee category i .

Use the z test statistic since σi's are known

H0: μ1 ≥ μ2 One-sided lower tailed test


H1: μ1 < μ2 ← non-attendees have lower job satisfaction than
attendees.

Region of Acceptance Use α = 0.05


z-crit = z(0.05) = -1.645

Decision rule Do not reject H0 in favour of H1 if -1.645 ≤ z-stat

z-stat std error = √(1.1²/22+0.8²/25)


0.283901

z-stat = ((6.9-7.5)-(0))/0.283901
-2.1134

Statistical conclusion
Since z-stat (-2.1134) lies outside (below) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that the mean job satisfaction score for non-attendees is significantly
lower than the mean job satisfaction score for job enrichment attendees.

Thus, the statistical evidence supports the view that the job enrichment
workshops significantly increased job satisfaction levels of sales consultants.
Exercise 9.11

(a) 95% Confidence Limits - Explorer Fund only


Std error = √(2.3²/15) = 0.5939
z (0.95) = 1.96

Lower 95% confidence limit = 12.4 - 1.96 (0.5939) = 11.24 days


Upper 95% confidence limit = 12.4 + 1.96 (0.5939) = 13.56 days

There is a 95% chance that the true average time to settlement for
claims lodged against the Explorer Fund lies between 11.24 days and 13.56 days.

(b) Let population 1 = Green-Aid Medical Fund


Let population 2 = Explorer Medical Fund
Let μi = population mean time to settlement of claims by each medical fund i .

Use the z test statistic since σi's are known

H0: μ1 ≥ μ2 One-sided lower tailed test


H1: μ1 < μ2 ← Green-Aid Fund settles quicker than Explorer Fund

Region of Acceptance (Use α = 0.05)


z-crit = z(0.05) = -1.645

Decision rule Do not reject H0 in favour of H1 if -1.645 ≤ z-stat

z-stat std error = √(3.2²/14+2.3²/15)


1.041199

z-stat = ((10.8-12.4)-(0))/1.041199
-1.5367
Statistical conclusion
Since z-stat (-1.5367) lies within the region of acceptance, there is insufficient sample
(or p-value = 0.0622 > α = 0.05) (see (iii) below), there is insufficient sample evidence
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
There is no difference in the mean claims settlement time between the two Funds.

Thus, the statistical evidence does not support the view that the Green-Aid Medical
Fund settles claims sooner, on average, than the Explorer Medical Fund.
Exercise 9.12

Let population 1 = Gas ovens


Let population 2 = Electric ovens
Let μi = population mean baking time for each oven type i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≥ μ2 One sided lower tailed test


H1: μ1 < μ2 ← gas ovens bake faster than electric ovens

Region of Acceptance Use α = 0.05 with df = (5+5-2) = 8


t-crit = t(0.05)(8) = -1.86

Decision rule Do not reject H0 in favour of H1 if t-stat ≥ -1.86

t-stat s² (pooled variance) = ((5-1)0.16²+(5-1)0.09²)/(5+5-2)


0.01685

std error = √(0.01685*(1/5+1/5))


0.08210

t-stat = ((0.75-0.89)-(0))/0.0821
-1.705

Statistical conclusion
Since t-stat (-1.705) lies within the region of acceptance, there is insufficient sample
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that there is no difference in the mean bread baking time between gas
and electric ovens, at the 5% significance level.
Gas ovens are therefore, not faster, on average, than electric ovens.
Exercise 9.13

(a) Test for equality of variances


H0: σ21 = σ22 H1: σ²1 ≠ σ²2

Rule of thumb test: F-stat < 3?


F-stat = = 152.2²/121.5² = 1.569
Since F-stat (1.569) < 3, do not reject H0 in favour of H1.
Conclude that the two population variances are equal.
Therefore, use the pooled-variances t -test to test for differences in means.

(b) Let population 1 = Cape Town branch


Let population 2 = Durban branch
Let μi = population mean size of orders received by each branch i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≤ μ2 One sided upper tailed test


H1: μ1 > μ2 ← CT branch performing better than Durban branch

Region of Acceptance Use α = 0.10 with df = (18+15-2) = 31


t-crit = t(0.10)(31) = 1.309

Decision rule Do not reject H0 in favour of H1 if t-stat ≤ 1.309

t-stat s² (pooled variance) = ((18-1)121.5²+(15-1)152.2²)/(18+15-2)


18556.97

std error = √(18556.97*(1/18+1/15))


47.6243

t-stat = ((335.2-265.6)-(0))/47.6243
1.4614
Statistical conclusion
Since t-stat (1.4614) lies outside (above) the region of acceptance, there is sufficient
sample evidence at the 10% significance level to reject H0 in favour of H1 (i.e. Reject H0).

Management conclusion
Conclude that the mean size of orders received by the Cape Town branch is
significantly larger than the mean size of orders received by the Durban branch.

Thus the Cape Town branch is performing better than the Durban branch in
terms of average order size.

(c) Region of Acceptance (new) Use α = 0.05 with df = (18+15-2) = 31


t-crit = t(0.05)(31) = 1.696
Decision rule (new) Do not reject H0 in favour of H1 if t-stat ≤ 1.696

Statistical conclusion
t-stat (= 1.4614) now lies within the (new) region of acceptance.
Thus there is insufficient sample evidence at the 5% level of significance to
reject H0 in favour of H1. (i.e. Accept H0).
Management conclusion
Conclude that there is no significant difference in the mean order sizes between the
Cape Town branch and the Durban branch.

Thus there is no evidence to believe that the Cape Town branch is performing better
than the Durban branch in terms of average order size.

(d) Findings based on the 5% significance level are more meaningful than those based
on the 10% significance level because it requires stronger (more convincing) sample

evidence before tests conducted at 5% are prepared to reject the null hypothesis.

The operations manager can be more confident that there is no difference in mean
performance between the two branches (conclusion based on (b)).
Exercise 9.14 File: X9.14 - package designs.xlsx

First, test for equality of variances

H0: σ21 = σ22 Two tailed test


H1: σ²1 ≠ σ²2

F-cri t = F(0.05/2,7,7) = 4.99 (from Table 4(b), Appendix 1) =F.INV.RT(0.025,7,7)


Decision rule: Do not reject H0 if F-stat ≤ 4.99 4.995

F-stat = = 5.706²/4.862² = 1.377 Use rule of larger variance in numerator

Since F-stat (1.377) < F-crit (4.99), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Therefore, use the pooled-variances t -test to test for differences in means.

Now conduct the t-test for equal means, using the pooled-variances t-test approach.

Let population 1 = Pyramid-shaped carton


Let population 2 = Barrel-shaped carton
Let μi = population mean sales volume of one-litre cartons for each carton shape i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≥ μ2 One sided lower tailed test


H1: μ1 < μ2 'Pyramid' sales are less than 'Barrel' sales

Region of Acceptance Use α = 0.05 with df = (8+8-2) = 14


t-crit = t(0.05)(14) = -1.761

Decision rule Do not reject H0 in favour of H1 if -1.761 ≤ t-stat

t-stat s² (pooled variance) = ((8-1)4.862²+(8-1)5.706²)/(8+8-2)


28.09874

std error = √(28.09874*(1/8+1/8))


2.650412

t-stat = ((23.75-27.375)-(0))/2.650412
-1.3677

Statistical conclusion
Since t-stat (-1.3677) lies within the region of acceptance, there is insufficient sample
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that there is no difference in the mean weekly sales of one-litre cartons
of apple juice between the pyramid-shaped and barrel-shaped carton designs.

Thus the marketer can choose either package design to achieve higher weekly sales.
Exercise 9.15

Let population 1 = Fruit Puffs consumers


Let population 2 = Fruity Wheat consumers
Let πi = population proportion of consumers who prefer fruit-flavoured wheat cereal i .

Use the z test statistic

H0: π1 ≤ π2 One sided upper tailed test


H1: π1 > π2 Fruit Puffs is preferred by more consumers than
Fruity Wheat.

Region of Acceptance Use α = 0.05


z-crit = z(0.05) = 1.645

Decision rule Do not reject H0 in favour of H1 if z-stat ≤ 1.645.

Sample data Fruit Puffs Fruity Wheat


n 175 150
x 54 36
pi 0.309 0.240

z-stat π(hat) (pooled proportion) = (54+36)/(175+150) =


0.2769

std error = √(0.2769*(1-0.2769)*(1/175+1/150))


0.04978949

z-stat = (0.309-0.24)/0.049789
1.3858

Statistical conclusion
Since z-stat (= 1.3858) lies within the region of acceptance, there is insufficient sample
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that there is no difference in the percentage of consumers who prefer
each type of fruit-flavoured wheat cereal.

The marketer's view that Fruit Puffs is more preferred than Fruity Wheat cannot be
validated based on the statistical evidence at the 5% level of significance.
The marketer can therefore choose to launch either fruit flavour of wheat cereal.
Exercise 9.16

(a) Let population 1 = Male respondents


Let population 2 = Female respondents
Let πi = population proportion who prefer jazz for each gender i .

Use the z test statistic

H0: π1 = π2 Two sided test (equal preference)


H1: π1 ≠ π2

Region of Acceptance Use α = 0.05


z-crit = 1.96

Decision rule Do not reject H0 in favour of H1 if -1.96 ≤ z-stat ≤ 1.96.

Sample data Male Female


n 140 110
x 46 21
pi 0.329 0.191

z-stat π(hat) (pooled proportion) = (46+21)/(140+110) =


0.268

std error = √(0.268*(1-0.268)*(1/140+1/110))


0.056432928

z-stat = (0.329-0.191)/0.056433
2.445

Statistical conclusion
Since z-stat (2.445) lies outside (above) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1 (i.e. Reject H0)

Management conclusion
Conclude that there is a difference in the proportion of males compared to the
proportion of females who enjoy listening to jazz.

By inspection , proportionately more males than females enjoy listening to jazz.

(b) p -value = =(1-NORMSDIST(2.445))*2 0.0145

Since the p-value (=0.0145) < α = 0.05, there is strong sample evidence in
support of H1. Hence conclude that there is a difference in the proportion of males
compared to the proportion of females who enjoy listening to jazz.
Exercise 9.17

(a) Let population 1 = Status Cheque Account clients


Let population 2 = Elite Cheque Account clients
Let πi = population proportion of clients for each account type i who are overdrawn.

Use the z test statistic

H0: π1 ≥ π2 One sided lower tailed test


H1: π1 < π2 'Status' proportion less than 'Elite' proportion

Region of Acceptance Use α = 0.05


z-crit = -1.645

Decision rule Do not reject H0 in favour of H1 if z-stat ≥ -1.645.

Sample data Status Elite


n 300 250
x 48 55
pi 0.16 0.22

z-stat π(hat) (pooled proportion) = (48+55)/(300+250) =


0.1873

std error = √(0.1873*(1-0.1873)*(1/300+1/250))


0.0334

z-stat = (0.16-0.22)/0.0334
-1.796
Statistical conclusion
Since z-stat (= -1.796) lies outside (below) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that proportionately more Elite cheque account clients are overdrawn
compared to Status cheque account clients.

(b) p -value = =NORMSDIST(-1.796) 0.0363

Since the p-value (=0.0363) < α = 0.05, there is strong sample evidence in
support of H1. Hence conclude that proportionately more Elite cheque account clients
are overdrawn compared to Status cheque account clients.
Exercise 9.18 File: X9.18 - aluminium scrap.xlsx

(a) (i) Test for equality of variances

2 2
H0: σ 1 =σ2 Two tailed test
H1: σ²1 ≠ σ²2

Since F-stat (1.702) < F-crit (1.77), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Use the pooled-variances t -test to test for differences in means.

F-Test Two-Sample for Variances (Data Analysis - Excel )


Machine 1 Machine 2
Mean 3.483 3.668
Variance 0.261 0.153
Observations 50 30
df 49 29
F-stat 1.702
P(F<=f) one-tail 0.0639
F Critical one-tail 1.777

(a) (ii) Let population 1 = machine 1 daily % scrap


Let population 2 = machine 2 daily % scrap
Let μi = population average daily % scrap produced by each machine i .

(a) (iii) Use the pooled-variances t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≥ μ2 One sided lower tailed test


H1: μ1 < μ2 Machine 1 produces lower % scrap than machine 2

Test performed manually using Descriptive Statistics from Data Analysis

Region of Acceptance Use α = 0.05 with df = (50+30-2) = 78


t-crit = t(0.05)(78) = TINV(0.1,78) -1.664

Decision rule Do not reject H0 in favour of H1 if -1.664 ≤ t-stat

Sample data - descriptive statistics


Machine 1 Machine 2
Mean 3.483 3.668
Standard Error 0.072 0.072
Median 3.41 3.705
Mode 2.97 3.77
Standard Deviation 0.511 0.392
Sample Variance 0.261 0.153
Skewness 0.331 0.014
Range 1.83 1.59
Minimum 2.67 2.88
Maximum 4.5 4.47
Sum 174.14 110.03
Count 50 30
t-stat s² (pooled variance) = ((50-1)*0.511²+(30-1)*0.392²)/(50+30-2)
0.221169038
std error = √(0.221169*(1/50+1/30))
0.108607919

t-stat = ((3.483-3.668)-(0))/0.10861
-1.7032

(a) (iii) Test performed using t-Test Assuming Equal Variances in Data Analysis

Using Data Analysis (in Excel )


t-Test: Two-Sample Assuming Equal Variances

Machine 1 Machine 2
Mean 3.483 3.668
Variance 0.261 0.153
Observations 50 30
Pooled Variance 0.221
Hypothesized Mean Difference 0
df 78
t Stat -1.703
P(T<=t) one-tail 0.046
t Critical one-tail -1.665
P(T<=t) two-tail 0.093
t Critical two-tail 1.991

(a) (iv) Statistical conclusion


Since t-stat (= -1.7032) lies outside (below) the region of acceptance, there is sufficient
or p-value = 0.0463 < α = 0.05 (see below), there is sufficient sample evidence
sample evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that machine 1 has a significantly lower average daily % scrap than
machine 2.

(b) p -value = =T.DIST(-(-1.7032),78,1) 0.0463 =T.DIST(-1.7032,78,TRUE)

Since the p-value (=0.046) < α = 0.05, there is strong sample evidence in
support of H1. Conclude that machine 1 produces scrap at a significantly lower
average daily rate than machine 2. This conclusion is valid at the 5% significance level.
Exercise 9.19 File: X9.19 - water purification.xlsx

(a) (i) Test for equality of variances

2 2
H0: σ 1 =σ 2 Two tailed test
H1: σ²1 ≠ σ²2

Since F-stat (1.063) < F-crit (1.924), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Use the pooled-variances t -test to test for differences in means.

F-Test Two-Sample for Variances (Data Analysis - Excel )


Free State KZN
Mean 27.458 26.448
Variance 1.563 1.470
Observations 24 29
df 23 28
F-stat 1.063
P(F<=f) one-tail 0.4342
F Critical one-tail 1.924

(a) (ii) Let population 1 = Free State plant daily impurities levels
Let population 2 = KZN plant daily impurities levels
Let μi = population average daily impurities level per plant i .

(a) (iii) Use the pooled-variances t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≤ μ2 One sided upper tailed test


H1: μ1 > μ2 FS treatment plant has higher level of impurities
than the KZN water treatment plant.

Test performed manually using Descriptive Statistics from Data Analysis

Region of Acceptance Use α = 0.01 with df = (24+29-2) = 51


t-crit = t(0.01)(51) = TINV(0.02,51) 2.402

Decision rule Do not reject H0 in favour of H1 if t-stat ≤ 2.402

Descriptive Statistics Free State KZN


Mean 27.458 26.448
Standard Error 0.255 0.225
Median 27.5 27
Mode 28 27
Standard Deviation 1.250 1.213
Sample Variance 1.563 1.470
Skewness 0.030 -0.193
Range 5 5
Minimum 25 24
Maximum 30 29
Sum 659 767
Count 24 29
Confidence Level(99.0%) 0.717 0.622

t-stat s² (pooled variance) = ((24-1)*1.25²+(29-1)*1.213²)/(24+29-2)


1.51247
std error = √(1.51247*(1/24+1/29))
0.33937

t-stat = ((27.458-26.448)-(0))/0.33937
2.9764
Test performed using t-Test Assuming Equal Variances in Data Analysis

Using Data Analysis (in Excel )


t-Test: Two-Sample Assuming Equal Variances

Free State KZN


Mean 27.458 26.448
Variance 1.563 1.470
Observations 24 29
Pooled Variance 1.512
Hypothesized Mean Difference 0
df 51
t Stat 2.9764
P(T<=t) one-tail 0.00223
t Critical one-tail 2.402
P(T<=t) two-tail 0.004
t Critical two-tail 2.676

(a) (iv) Statistical conclusion


Since t-stat (2.9764) lies outside (above) the region of acceptance, there is sufficient
sample evidence at the 1% significance level to reject H 0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that the KZN plant produces water of a higher quality (fewer impurities, on average)
than the Free State plant. Thus the KZN plant manager's claim can be supported
statistically at the 1% level of significance.

(b) p -value = =T.DIST(2.9761),51,1) 0.00223 =T.DIST.RT(2.9764,51)

Since the p-value (=0.00223) << α = 0.01, there is overwhelming sample evidence in
support of H1. Same conclusion as in (a) applies.
Exercise 9.20 File: X9.20 - herbal tea.xlsx

(a) Test for equality of variances

H0: σ21 = σ22 Two tailed test


H1: σ²1 ≠ σ²2

Since F-stat (1.227) < F-crit (2.098), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Use the pooled-variances t -test to test for differences in means.

F-Test Two-Sample for Variances (Data Analysis - Excel )


Freshpak Yellow Label
Mean 7.689 6.917
Variance 2.170 1.768
Observations 19 23
df 18 22
F-stat 1.227
P(F<=f) one-tail 0.3205
F Critical one-tail 2.098

(b) (i) Let population 1 = Freshpak brand


Let population 2 = Yellow Label brand
Let μi = population mean level of quercetin in mg/kg in each brand i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 = μ2 Two sided test (No difference)


H1: μ1 ≠ μ2

Test performed manually using Descriptive Statistics from Data Analysis

Region of Acceptance Use α = 0.05 with df = (19+23-2) = 40


t-crit = t(0.05)(40) = 2.021

Decision rule Do not reject H0 in favour of H1 if -2.021 ≤ t-stat ≤ 2.021

Descriptive Statistics Freshpak Yellow Label


Mean 7.689 6.917
Standard Error 0.3379 0.2772
Median 7.9 7.1
Mode 7.9 5.7
Standard Deviation 1.4731 1.3296
Sample Variance 2.1699 1.7679
Skewness -0.1752 0.1030
Range 5.1 5.3
Minimum 5 4.5
Maximum 10.1 9.8
Sum 146.1 159.1
Count 19 23
t-stat s² (pooled variance) = ((19-1)*1.473²+(23-1)*1.3296²)/(19+23-2)
1.948688

std error = √(1.948688*(1/19+1/23))


0.43277

t-stat = ((7.689-6.917)-(0))/0.43277
1.784

Test performed using t-Test Assuming Equal Variances in Data Analysis

Using Data Analysis (in Excel )


t-Test: Two-Sample Assuming Equal Variances

Freshpak Yellow Label


Mean 7.6895 6.9174
Variance 2.1699 1.7679
Observations 19 23
Pooled Variance 1.9488
Hypothesized Mean Difference 0
df 40
t Stat 1.7840
P(T<=t) one-tail 0.0410
t Critical one-tail 1.6839
P(T<=t) two-tail 0.0820
t Critical two-tail 2.0211

Statistical conclusion
Since t-stat (1.784) lies within the region of acceptance, there is insufficient sample
evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that there is no difference in the mean quercetin content (in mg/kg)
between the Freshpak and Yellow Labels brands of rooibos tea.
(b) (ii) Let population 1 = Freshpak brand (FP)
Let population 2 = Yellow Label brand (YL)

H0: μ1 ≤ μ2 One sided upper tailed test


H1: μ1 > μ2 FP contains more quercetin than YL

Using Data Analysis (in Excel )


t-Test: Two-Sample Assuming Equal Variances

Freshpak Yellow Label


Mean 7.6895 6.9174
Variance 2.1699 1.7679
Observations 19 23
Pooled Variance 1.9488
Hypothesized Mean Difference 0
df 40
t Stat 1.7840
P(T<=t) one-tail 0.0410
t Critical one-tail 1.6839
P(T<=t) two-tail 0.0820
t Critical two-tail 2.0211

Statistical conclusion
Since t-stat (= 1.784) lies outside (above) the region of acceptance (t-stat ≤ 1.6839),
(see table above), there is sufficient sample evidence at the 5% level of
significance to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that Freshpak's claim that their brand contains more quercetin, on average,
than the Yellow Label brand, can be supported statistically at the 5% significance level.
Exercise 9.21 File: X9.21 - meat fat.xlsx

(a) (i) Test for equality of variances

H0: σ21 = σ22 Two tailed test


H1: σ²1 ≠ σ²2

Since F-stat (1.261) < F-crit (2.066), do not reject H0 in favour of H1.

Conclude,at α = 0.05, that the two population variances are equal.


Use the pooled-variances t -test to test for differences in means.

F-Test Two-Sample for Variances (Data Analysis - Excel )


Namibia Little Karoo
Mean 27.074 30.333
Variance 33.840 26.833
Observations 27 21
df 26 20
F 1.261
P(F<=f) one-tail 0.3002
F Critical one-tail 2.066

(a) (ii) Let population 1 = Namibian meat producer


Let population 2 = Little Karoo meat producer
Let μi = population average fat content of meat supplied by each producer i .

Use the t test statistic since σi's are unknown (only s1 and s2 are given)

H0: μ1 ≥ μ2 One sided lower tailed test


H1: μ1 < μ2

(a) (iii) Test performed manually using Descriptive Statistics from Data Analysis

Region of Acceptance Use α = 0.01 with df = (27+21-2) = 46


t-crit = t(0.01)(46) = =TINV(0.02,46) 2.412

Decision rule Do not reject H0 in favour of H1 if t-stat ≤ 2.412

t-stat s² (pooled variance) =((27-1)*5.8173²+(21-1)*5.1801²)/(27+21-2)


30.794

std error = √(30.794*(1/27+1/21))


1.61459

t-stat = ((27.0741-30.3333)-(0))/1.61459
-2.0186
Test performed using t-Test Assuming Equal Variances in Data Analysis

Using Data Analysis (in Excel )


t-Test: Two-Sample Assuming Equal Variances

Namibia Little Karoo


Mean 27.0741 30.3333
Variance 33.8405 26.8333
Observations 27 21
Pooled Variance 30.7939
Hypothesized Mean Difference 0
df 46
t Stat -2.0186
P(T<=t) one-tail 0.0247
t Critical one-tail -2.4102
P(T<=t) two-tail 0.0494
t Critical two-tail 2.6870

(a) (iv) Statistical conclusion


Since t-stat (-2.0186) lies within the region of acceptance, there is insufficient sample

evidence at the 1% significance level to reject H0 in favour of H1. (i.e. Accept H0).

Management conclusion
Conclude that the mean fat content of meat between the Namibian producer and the
Little Karoo producer is the same .

There is therefore no statistical justification, at the 1% significance level, to sign an


exclusive agreement with the Namibian producer.

(b) p -value = =TDIST(-(-2.0186),46,1) 0.0247 =T.DIST(-2.0186,46,TRUE)


This is the same p-value as shown in the Data Analysis output for a one-tailed test.

Since p-value = 0.0247 > α = 0.01 (see Table above), there is insufficient sample evidence
to reject H0 in favour of H1 at the 1% level of signficance.
Same management conclusion as in (a) (iii) above.
Exercise 9.22 File: X9.22 - disinfectant sales.xlsx

(a) Matched pairs test


The same retail outlets were surveyed both before and after the
promotional campaign. Thus the two samples are not independent.

(b) Define x1 = Sales per outlet before the promotional campaign


x2 = Sales per outlet after the promotional campaign

Let d = (x1 - x2) i.e. " before" - "after"


Let μd = population mean difference in sales from before to after the promotional campaign.

Use the matched pairs t test statistic

H0: μd ≥ 0 One sided lower tailed test


H1: μd < 0 'before' sales are lower than the 'after' sales

(c) Region of Acceptance Use α = 0.05 with df = (12-1) = 11


t-crit = t(0.05)(11) = -1.796

Decision rule Do not reject H0 in favour of H1 if -1.796 ≤ t-stat

Sample data Σxd -8


n 12
x(bar)d -0.667
sd 1.231

t-stat (-0.667 - 0)/(1.231/√12) =


-1.877

Statistical conclusion
Since t-stat (-1.877) lies outside (below) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Reject H0).

Management conclusion
Conclude that there has been a significant increase in mean sales of 500ml bottles
of disinfectant liquid from before to after the promotional campaign.

The promotional campaign has therefore been a success at significantly


increasing mean sales volume of the product - at the 5% significant level.
(d) t-Test: Paired Two Sample for Means using Data Analysis

Before After
Mean 11.5 12.17
Variance 4.455 3.606
Observations 12 12
Pearson Correlation 0.817
Hypothesized Mean Difference 0
df 11
t Stat -1.876
P(T<=t) one-tail 0.0437
t Critical one-tail 1.7959
P(T<=t) two-tail 0.0874
t Critical two-tail 2.2010

p- value = =TDIST(-(-1.877),11,1) 0.0436

Statistical conclusion
Since p -value = 0.0437 < α = 0.05, there is moderately strong sample
evidence to reject H0 in favour of H1 and conclude that the promotional
campaign has been effective.
Exercise 9.23 File: X9.23 - performance ratings.xlsx

(a) The samples are dependent as the same employee is tested both before and
after the training sessions.

(b) Matched pairs test


x1 = Performance rating before the training sessions
x2 = Performance rating after the training sessions

Let d = (x1 - x2) i.e. " before" - "after"


Let μd = population mean difference in performance rating scores from
before to after the training sessions.

Use the matched pairs t test statistic

H0: μd ≥ 0 One sided lower tailed test


H1: μd < 0 'before rating scores lower than 'after' rating scores.

Test performed manually

Region of Acceptance Use α = 0.05 with df = (18-1) = 17


t-crit = t(0.05)(17) = -1.74

Decision rule Do not reject H0 in favour of H1 if -1.74 ≤ t-stat

Sample data Σxd -6.4


n 18
x(bar)d -0.356
sd 0.7114

t-stat (-0.356 - 0)/(0.7114/√18) =


-2.123

Data Analysis: t-Test : Paired Two Sample for Means

Before After
Mean 11.067 11.422
Variance 6.024 5.835
Observations 18 18
Pearson Correlation 0.95743865
Hypothesized Mean Difference 0
df 17
t Stat -2.120
P(T<=t) one-tail 0.0245
t Critical one-tail 1.740
P(T<=t) two-tail 0.04898861
t Critical two-tail 2.10981556

Statistical conclusion
Since t-stat (-2.123) lies outside (below) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1 (i.e. Reject H0)
Management conclusion
Conclude that there has been a significant increase in mean performance ratings
scores of employees who attended the series of workshops ans seminars.

The performance enhancement sessions have therefore been effective at increasing


motivation and productivity - at the 5% level of significance.

(c) p -value = =TDIST(-(-2.123),17,1) 0.0244 Also see t-Test Table above.

Since p-value = 0.0244 < α = 0.05, there is strong sample evidence to reject H0
in favour of H1 and conclude that the workshops have significantly increased
employee motivation and productivity.
Exercise 9.24 File: X9.24 - household debt.xlsx

(a) The samples are dependent as the same household is tested both a year ago
and at the current time period .

(b) Matched pairs test


x1 = Household debt level a year ago.
x2 = Household debt level currently.

Let d = (x1 - x2) i.e. " year ago" - "current"


Let μd = population mean difference in household debt levels from a year ago
to the current period.

Use the matched pairs t test statistic

H0: μd ≤ 0 One sided upper tailed test


H1: μd > 0 Debt higher a year ago than today (current period)

Test performed manually

Region of Acceptance Use α = 0.05 with df = (10-1) = 9


t-crit = t(0.05)(9) = 1.833

Decision rule Do not reject H0 in favour of H1 if t-stat ≤ 1.833

Sample data Σxd 12


n 10
x(bar)d 1.2
sd 1.8135

t-stat (1.2 - 0)/(1.8135/√10) =


2.0925

Data Analysis: t-Test : Paired Two Sample for Means

Year ago Current


Mean 40.5 39.3
Variance 35.611 36.011
Observations 10 10
Pearson Correlation 0.9541
Hypothesized Mean Difference 0
df 9
t Stat 2.0925
P(T<=t) one-tail 0.0330
t Critical one-tail 1.8331
P(T<=t) two-tail 0.0659
t Critical two-tail 2.2622

Statistical conclusion
Since t-stat (2.0925) lies outside (above) the region of acceptance, there is sufficient
sample evidence at the 5% significance level to reject H0 in favour of H1. (i.e. Reject H0).
Management conclusion
Conclude that there has been a significant decrease in the average level of
household debt from a year ago.

The increase in prime interest rate (from 6% to 11%) has lead to a significant decline
in the average level of household debt from a year ago - at the 5% significance level.

(c) p -value = =TDIST(2.0925,9,1) 0.033 Also see t-Test Table above

Since the p -value (0.033) < α = 0,05, there is strong sample evidence to reject H0 in
favour of H1 at the 5% level of significance.
Same statistical and management conclusions as (b) above.
Exercise 9.25 Process output variability study

Random variable: Hourly output per process


unit of measure: units produced per hour

2 2
H0 σ (1) = σ (2) Management question in H0
H1: 2
σ (1) ≠ σ (2) 2

Region of Acceptance (use α = 0.05)


Note: Set up the F-test as an upper tailed test (F-stat = larger s2/smaller s2)

F-crit (upper) = F(0.05,30,24) 1.939


Decision rule: Do not reject H0 if F-stat (the sample evidence) ≤ 1.939

Sample data α= 0.05


s12 = 14.6 s22 = 23.2
n1 = 25 n2 = 31

F-stat = larger s2/smaller s2 F-stat = 23.2/14.6 = 1.589


p -value = 0.1240
Statistical conclusion
Since F-stat = 1.589 < F-crit = 1.939, there is insufficient sample evidence at the
5% level of significance to reject H0 in favour of H1.
Management conclusion
Therefore conclude, with 95% confidence, that the variability of hourly outputs
between the two production processes is the same.
Exercise 9.26 File: X9.26 - milk yield.xlsx

Random variable: milk yield (in litres per week) per cow

H0: σ2(f) ≤ σ2(c)


H1: σ2(f) > σ2(c) Management question in H1

Region of Acceptance: Use α = 0.05 with df1 = 16-1 = 15 and df2 = 16-1 = 15
F-crit = F(0.05,15,15) = 2.403 (See Excel output)
Decision rule: Do not reject H0 if F-stat ≤ 2.403

F-Test Two-Sample for Variances (Data Analysis in Excel )


Free grazing Controlled Feed
Mean 31.306 35.375
Variance 72.138 43.110
Observations 16 16
df 15 15
F-stat 1.673
P(F<=f) one-tail 0.1647
F Critical one-tail 2.403

F-stat = 1.673 (See Excel output) p -value = 0.1647

Statistical conclusion
Since F-stat = 1.673 < F-crit = 2.403, there is insufficient sample evidence at the
5% level of significance to reject H0 in favour of H1.
Management conclusion
Therefore conclude, with 95% confidence, that there is no signficant difference
in the variability in milk yields of cows between the two feeding practices.
Exercise 9.27 File: X9.27 - employee wellness.xlsx

Random variable: Hours spent exercising per week

H0: σ2(over 40) ≤ σ2(under 40)


H1: σ2(over 40) > σ2(under 40) Management question in H1

Region of Acceptance: Use α = 0.05 with df1 = 23-1 = 22 and df2 = 21-1 = 20
F-crit = F(0.05,22,20) = 2.102 (See Excel output)
Decision rule: Do not reject H0 if F-stat ≤ 2.102

F-Test Two-Sample for Variances (Data Analysis (Excel ))


Over 40 Under 40
Mean 2.461 3.086
Variance 0.836 0.373
Observations 23 21
df 22 20
F-stat 2.240
P(F<=f) one-tail 0.0373
F Critical one-tail 2.102

F-stat = 2.24 (See Excel output) p -value = 0.0373

Statistical conclusion
Since F-stat = 2.24 > F-crit = 2.102, there is sufficient sample evidence at the
5% level of significance to reject H0 in favour of H1.
Management conclusion
Therefore conclude, with 95% confidence, that 'over 40' employees do indeed
exercise more 'erractically' (i.e. show significantly greater variability in exercise times)
than 'under 40' employees.
Exercise 9.28 File: X9.28 - attrition rate.xlsx

Random variable Attrition rates (%) per call center per month

2 2
H0: σ (fin) = σ (health) Management question in H0
2 2
H1: σ (fin) ≠ σ (health)

Region of Acceptance of H0: Find only F-crit (lower) or F-crit (upper)


This depends on whether the smaller or the larger sample variance is in the numerator of F-stat

For an F-crit (upper) only: For an F-crit (lower) only:


F-stat = Larger variance / Smaller variance F-stat = Smaller variance / Larger variance

F-Test Two-Sample for Variances F-Test Two-Sample for Variances


Health Financial Financial Health
Mean 5.46 6.13 Mean 6.13 5.46
Variance 1.1413 0.6653 Variance 0.6653 1.1413
Observations 17 21 Observations 21 17
df 16 20 df 20 16
F-stat 1.715 F-stat 0.583
P(F<=f) one-tail 0.1263 P(F<=f) one-tail 0.1263
F-crit (upper) 2.184 F-crit (lower) 0.458

F-crit (upper) = F(0.05,16,20) =1.1413/0.6653 2.184


OR F-crit (lower) = 1/F(0.05,16,20) =0.6653/1.1413 0.458
(or F(0.95,20,16))
Decision rule: For upper tailed test: Do not reject H0 if F-stat ≤ 2.184
OR For lower tailed test: Do not reject H0 if F-stat ≥ 0.458

Now F-stat = 1.715 (for an upper tailed test); or F-stat = 0.583 (for a lower tailed test)

Conclusion (based on an upper tailed F-test)


Since F-stat = 1.715 > F-crit (upper) = 2.184 (and hence lies within the acceptance region), there
is insufficient sample evidence at the 5% level of significance to reject H0 in favour of H1.
Same Conclusion (based on a lower tailed F-test)
Since F-stat = 0.583 > F-crit (lower) = 0.458 (and hence lies within the acceptance region), there
is insufficient sample evidence at the 5% level of significance to reject H0 in favour of H1.

Management conclusion
With 95% confidence, it can be concluded that there is no significant difference in the variability
in attrition rates between the two sectors (financial and health).
CHAPTER 10

CHI-SQUARE HYPOTHESIS TESTS

Exercise 10.1 Purpose: To test whether there is any statistically significant association between
the outcomes of two categorical variables. Stated differently, are the outcomes
associated with two categorical variables independent of each other or not?

Exercise 10.2 Categorical (nominal or ordinal-scaled) data

Exercise 10.3 H0: There is no statistical association between the two categorical variables

Exercise 10.4 Expected frequencies represent the null hypothesis of no association (or statistical
independence ) between the two categorical variables.

Exercise 10.5 χ²-crit (0.05,6) = 12.592 χ²-crit (0.10,6) = 10.645


Exercise 10.6 File: X10.6 - motivation status.xlsx

(a) Row Percentages

Motivation level
High Moderate Low Total
Male 26.7 26.7 46.7 100
Female 47.5 30.0 22.5 100
Total 38.6 28.6 32.9 100

Interpretation (by inspection)


When compared to the general population profile, males tend to have low levels of
motivation, while females tend to be more highly motivated. It appears
therefore that a statistical association exists between gender and motivation level.

(b) H0: There is no association between Gender and Motivation level


H1: There is an association between Gender and Motivation level

Region of Rejection (Use α = 0.10 with degrees of freedom = (2-1)(3-1) = 2)


χ²-crit = χ²(0.10)(2) = 4.605
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 4.605

χ²-stat Observed frequencies (fo)


High Moderate Low Total
Male 8 8 14 30
Female 19 12 9 40
Total 27 20 23 70

Expected frequencies (fe)


High Moderate Low Total
Male 11.57 8.57 9.86 30
Female 15.43 11.43 13.14 40
Total 27 20 23 70

Chi-Squared components
High Moderate Low
Male 1.102 0.038 1.741
Female 0.827 0.029 1.306
χ²-stat = 5.0428

Conclusion
Since χ²-stat = 5.0428 > χ²-crit = 4.605, there is sufficient sample evidence at the
10% significance level to reject H0 in favour of H1. Therefore conclude that there is a
statistical association between the gender of an employee and their level of motivation.
The nature of the relationship is described in (a) above.
Exercise 10.7 File: X10.7 - internet shopping.xlsx

(a) Row Percentages

Internet shopping
Yes No Total
full-time 24.3 75.7 100
at-home 18.5 81.5 100
Total 20.8 79.2 100

Interpretation (by inspection)


Since the row percentage profiles (between full-time employed and at-home customers) are
very similar to each other and to the general population profile, it can be concluded, by
observation, that the two attributes are not associated (i.e. they are statistically independent).

(b) H0: There is no association between Employment Status and Use of Internet Shopping
H1: There is an association between Employment Status and Use of Internet Shopping

Region of Rejection (Use α = 0.05 with degrees of freedom = (2-1)(2-1) = 1)


χ²-crit = χ²(0.05)(1) = 3.843
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 3.843

χ²-stat Observed frequencies (fo)


Yes No Total
full-time 35 109 144
at-home 40 176 216
Total 75 285 360

Expected frequencies (fe)


Yes No Total
full-time 30 114 144
at-home 45 171 216
Total 75 285 360

Chi-Squared components
Yes No
full-time 0.8333 0.2193
at-home 0.5556 0.1462
χ²-stat = 1.7544

Conclusion
Since χ²-stat = 1.7544 < χ²-crit = 3.843, there is insufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is
no statistical association between the employment status of a customer and their
use of the internet for shopping purposes.
These two events are statistically independent.
Exercise 10.8 File: X10.8 - car size.xlsx

(a) Row Percentages

Car sizes bought


Small Medium Large Total
Under 30 15.2 33.3 51.5 100
30 - 45 21.1 36.8 42.1 100
Over 45 37.5 29.2 33.3 100
Total 26.3 33.0 40.7 100

Interpretation (by inspection)


With reference to the general population profile (Total row %),
under 30's tend to prefer larger cars; 30-45 year age car buyers marginally tend
towards medium to large cars, while over 45's strongly tend to prefer smaller cars.

(b) H0: There is no association between Age of car buyer and Car Size bought.
H1: There is an association between Age of car buyer and Car Size bought.

Region of Rejection (Use α = 0.01 with degrees of freedom = (3-1)(3-1) = 4)


χ²-crit = χ²(0.01)(4) = 13.277
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 13.277

χ²-stat Observed frequencies (fo)


Small Medium Large Total
Under 30 10 22 34 66
30 - 45 24 42 48 114
Over 45 45 35 40 120
Total 79 99 122 300

Expected frequencies (fe)


Small Medium Large Total
Under 30 17.38 21.78 26.84 66
30 - 45 30.02 37.62 46.36 114
Over 45 31.6 39.6 48.8 120
Total 79 99 122 300

Chi-Squared components
Small Medium Large
Under 30 3.134 0.002 1.910
30 - 45 1.207 0.510 0.058
Over 45 5.682 0.534 1.587
χ²-stat = 14.6247

Conclusion
Since χ²-stat = 14.6247 > χ²-crit = 13.277, there is sufficient sample evidence at the
1% significance level to reject H0 in favour of H1. Therefore conclude that there is a
statistical association between the age of a car buyer and the size of car bought.
(c) Management Conclusion
The nature of the statistical relationship found in (b) is described in (a) above.
Under 30's tend to prefer larger cars; 30-45 year age car buyers marginally tend
towards medium to large cars, while over 45's strongly tend to prefer smaller cars.
Recommendation
Target larger cars to the younger market and smaller cars to the older market.
Exercise 10.9 File: X10.9 - sports readership.xlsx

(a) Let πi = proportion of people who read Sports News in each of the i regions.

H0: π1 = π2 = π3
H1: At least one πi is different (i = 1,2,3)

Sample proportions E Cape W Cape KZN


0.160 0.104 0.250

(b) Region of Rejection Use α = 0.01 with degrees of freedom = (2-1)(3-1) = 2


χ²-crit = χ²(0.01)(2) = 9.21
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 9.21

χ²-stat Observed frequencies (fo)


E Cape W Cape KZN Total
No 84 86 78 248
Yes 16 10 26 52
Total 100 96 104 300

Expected frequencies (fe)


E Cape W Cape KZN Total
No 82.67 79.36 85.97 248
Yes 17.33 16.64 18.03 52
Total 100 96 104 300

Chi-Squared components
E Cape W Cape KZN
No 0.0215 0.5556 0.7395
Yes 0.1026 2.6496 3.5267
χ²-stat = 7.5954

Conclusion
Since χ²-stat = 7.5954 < χ²-crit = 9.21, there is insufficient sample evidence at the
1% significance level to reject H0 in favour of H1. Therefore conclude that the
proportion of people who read Sports News is the same in each Geographical Region.
These two events are statistically independent.

(c) H0: There is no association between the propensity to read Sports News and Region
H1: There is an association between the propensity to read Sports News and Region
Exercise 10.10 File: X10.10 - gym activity.xlsx

(a) Row Percentages

Gym activity
Spinning Swimming Circuit Total
Male 42.35 22.35 35.29 100
Female 52.73 29.09 18.18 100
Total 46.43 25.00 28.57 100

Interpretation (by inspection)


When compared to the general population of gym goers, more females than males
tend to prefer spinning and swimming; while more males relative to females tend to
prefer mainly doing the circuit.
The evidence is however not strongly convincing.

(b) H0: There is no association between Gender and preferred Gym Activity
H1: There is an association between Gender and preferred Gym Activity

Region of Rejection Use α = 0.10 with degrees of freedom = (2-1)(3-1) = 2


χ²-crit = χ²(0.10)(2) = 4.605
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 4.605

χ²-stat Observed frequencies (fo)


Spinning Swimming Circuit Total
Male 36 19 30 85
Female 29 16 10 55
Total 65 35 40 140

Expected frequencies (fe)


Spinning Swimming Circuit Total
Male 39.46 21.25 24.29 85
Female 25.54 13.75 15.71 55
Total 65 35 40 140

Chi-Squared components
Spinning Swimming Circuit
Male 0.304 0.238 1.345
Female 0.470 0.368 2.078
χ²-stat = 4.803

Conclusion
Since χ²-stat = 4.803 > χ²-crit = 4.605 (marginally), there is sufficient sample evidence
at the 10% significance level to reject H0 in favour of H1. Therefore conclude that
gender and preferred gym activity are associated.
The nature of the relationship is described in (a) above.

(c) New rejection region Use α = 0.05 with degrees of freedom = (2-1)(3-1) = 2
χ²-crit = χ²(0.05)(2) = 5.991
Decision rule Reject H0 in favour of H1 if χ²-stat ≥ 5.991

New decision Do not reject H0 at the 5% significance level,


since χ²-stat = 4.803 < χ²-crit = 5.991.
New conclusion There is no statistical association between gender and gym activity
(i.e. they are statistically independent) at the 5% level of significant.

(d) Let πi = proportion of females who prefer each gym activity i (spinning, swimming, circuit)

H0: π1 = π2 = π3
H1: At least one πi is different (i = 1,2,3)

Sample proportions Spin Swim Circuit


0.446 0.457 0.250

Statistical conclusion
The same statistical conclusion applies as in (b) (i.e. a statistical association exists)
or stated differently: at least one population proportion is different.
Management conclusion
By an inspection of the row percentage table in (a), it can be concluded that
spinning and swimming are the most preferred gym activities of females,
while doing the circuit is the least preferred gym activity of females.
Exercise 10.11 File: X10.11 - supermarket visits.xlsx

(a) Categorical Frequency Table - Supermarket Visits

Visits Customers Percent Belief %


Daily 36 20.0 25
3 / 4 times 55 30.6 35
Twice 62 34.4 30
Once only 27 15.0 10
Total 180 100 100

Interpretation (by inspection)


Once-a-week visits are the least common shopping behaviour (only 15%).
Only one-in-five shoppers (20%) shop daily.
The most common shopping pattern is either 3/4 times a week (30.6%) or twice a week (34.4%).
These two shopping behaviours represent 65% of all the sampled shoppers.

(b) Goodness-of-fit test for an empirical distribution


H0: The frequency of store visits per week is as per the manager's belief.
H1: The frequency of store visits per week differs significantly from the manager's belief.

Region of Rejection (Use α = 0.05 with degrees of freedom = (4-1) = 3)


χ²-crit = χ²(0.05)(3) = 7.815
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 7.815

χ²-stat Customers
Visits fo %fe fe χ²-stat
Daily 36 25 45 1.8
3 / 4 times 55 35 63 1.016
Twice 62 30 54 1.185
Once only 27 10 18 4.5
Total 180 100 180 8.501

Conclusion
Since χ²-stat = 8.501 > χ²-crit = 7.815, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
the shopping frequency of customers differs significantly from the manager's belief.
The nature of the relationship is described in (a) above.

(c) Interpretation
The manager's belief is that customers shop more frequently during a week.
frequently than the manager believes. For example, the survey found that
only 51% shop more than 3 times per week, while the manager assumed that this
percentage was 60%. Similarly, more customers prefer to shop only once or twice a
week (49%) compared to the manager's belief that this percentage was only 40%).
These differences are however not strongly significantly different.
Exercise 10.12 File: X10.12 - equity portfolio.xlsx

Goodness-of-fit test for an empirical distribution

H0: There is no change in the equity portfolio mix between 2008 and 2012.
H1: There is a significant change in the equity portfolio mix between 2008 and 2012.

Region of Rejection (Use α = 0.05 with degrees of freedom = (4-1) = 3)


χ²-crit = χ²(0.05)(3) = 7.815
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 7.815

χ²-stat Equities
Equity fo (2012) Ratio fe fe (2008) χ²-stat
Mining 900 2 900 0
Industrial 1400 3 1350 1.852
Retail 400 1 450 5.556
Financial 1800 4 1800 0
Total 4500 10 4500 7.407

Conclusion
Since χ²-stat = 7.407 < χ²-crit = 7.815, there is insufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
there is no significant change in the equity portfolio mix of the investor
between 2008 and 2012.
The equity portfolio profile is essentially the same in 2012 as it was in 2008.
Exercise 10.13 File: X10.13 - payment method.xlsx

(a) Goodness-of-fit test for an empirical distribution

H0: There is no change in the payment method for electronic goods.


H1: There is a significant change in the payment method for electronic goods.

Region of Rejection Use α = 0.05 with degrees of freedom = (3-1) = 2


χ²-crit = χ²(0.05)(2) = 5.991
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 5.991

χ²-stat Payment Methods


Payment fo % fe fe χ²-stat
Cash 41 23 46 0.543
Debit Card 49 35 70 6.3
Credit Card 110 42 84 8.048
Total 200 100 200 14.891

Conclusion
Since χ²-stat = 14.891 >> χ²-crit = 5.991, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
there is a significant shift in payment practices from the past.

(b) Management Interpretation


There is a significant shift in payment practices for electronic goods.
There is more emphasis on credit card payment (55%) today than in the past (42%).
Exercise 10.14 File: XS10.14 - package sizes.xlsx

(a) Goodness-of-fit test for an empirical distribution

H0: Limpopo sales pattern follows the national sales pattern.


H1: Limpopo sales pattern does not follow the national sales pattern.

Region of Rejection (Use α = 0.05 with degrees of freedom = (3-1) = 2)


χ²-crit = χ²(0.05)(2) = 5.991
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 5.991

χ²-stat Package Size Sales


Package fo Ratio fe fe χ²-stat
Large 190 3 162 4.840
Midsize 250 5 270 1.481
Small 100 2 108 0.593
Total 540 10 540 6.914

Conclusion
Since χ²-stat = 6.914 > χ²-crit = 5.991, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
the Limpopo sales pattern of cereal package sizes differs significantly from the
national sales pattern of package sizes sold.

(b) Management Interpretation


The Limpopo sales patterns differs significantly from the national sales pattern
of package sizes sold. By an inspection of the Limpopo sales profile relative to the
national pattern, Limpopo tends to sell more large sized packages relative
to the national pattern.
Exercise 10.15 File: X10.15 - compensation plan.xlsx

(a) Column Percentages

Regions
Plan Cape Gauteng Free State KZN Total
Present 62 75.7 67.1 72.7 70.8
New 38 24.3 32.9 27.3 29.2
Total 100 100 100 100 100

Interpretation (by inspection)


When compared to the national population profile of all employees,
the present compensation plan enjoys more support in Gauteng, and the
least support in the Cape Province where the new plan is favoured more.
The evidence is however not overwhelming (i.e. the profile differences are not large)

(b) Let πi = proportion of sales staff in favour of the present payment plan in each province i .

H0: π1 = π2 = π3 = π4
H1: At least one πi is different (i = 1,2,3,4)

Sample proportions Cape Gauteng Free State KZN


0.62 0.757 0.671 0.727

(c) Region of Rejection Use α = 0.10 with degrees of freedom = (2-1)(4-1) = 3


χ²-crit = χ²(0.10)(3) = 6.251
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 6.251

χ²-stat Observed frequencies (fo)


Plan Cape Gauteng Free State KZN Total
Present 62 140 47 80 329
New 38 45 23 30 136
Total 100 185 70 110 465

Expected frequencies (fe)


Plan Cape Gauteng Free State KZN Total
Present 70.75 130.89 49.53 77.83 329
New 29.25 54.11 20.47 32.17 136
Total 100 185 70 110 465

Chi-Squared components
Plan Cape Gauteng Free State KZN
Present 1.0828 0.6337 0.1289 0.0606
New 2.6194 1.5330 0.3119 0.1466
χ²-stat = 6.5169

Conclusion
Since χ²-stat = 6.5169 > χ²-crit = 6.251 (marginally), there is sufficient sample evidence
at the 10% significance level to reject H0 in favour of H1. Therefore conclude that
the support for the present compensation plan is different in at least one of the provinces.
The nature of the relationship is described in (a) above.

(d) Formulate as a test for independence of association between payment plan and province.
H0: There is no association between payment plan preference and province.
H1: There is an association between payment plan preference and province.

(e) New rejection region Use α = 0.05 with degrees of freedom = (2-1)(4-1) = 3
χ²-crit = χ²(0.05)(3) = 7.815
Decision rule Reject H0 in favour of H1 if χ²-stat ≥ 7.815

New decision Do not reject H0 at the 5% significance level,


since χ²-stat = 6.5169 < χ²-crit = 7.815.

New conclusion
There is no statistical association between payment plan preference and province.
(i.e. they are statistically independent) at the 5% level of significant.
The sample evidence is not strong enough (i.e. the sample proportion differences are
not great enough) to reject H0 in favour of H1 at the 5% level of significance.
Exercise 10.16 File: X10.16 - tyre defects.xlsx

(a) Row Percentages

Nature of defective tyre


technical mechanical material Total
Morning 22.1 61.8 16.2 100
Afternoon 30.2 46.5 23.3 100
Night 42.6 36.8 20.6 100
Total 31.5 48.2 20.3 100

Interpretation (by inspection)


When compared to the total production of defective tyres (total row %),
tyre defects due to mechanical problems tend to be more prevalent during morning shifts.
Technical defects however tend to be more prevalent during the night shift.
Thus there does appear to be an association between shift and nature of tyre defects.

(b) Formulate as a test for independence of association between nature of defect and shift

H0: There is no association between nature of tyre defect and shift.


H1: There is an association between nature of tyre defect and shift.

Region of Rejection Use α = 0.05 with degrees of freedom = (3-1)(3-1) = 4


χ²-crit = χ²(0.05)(4) = 9.488
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 9.488

χ²-stat Observed frequencies (fo)


technical mechanical material Total
Morning 15 42 11 68
Afternoon 26 40 20 86
Night 29 25 14 68
Total 70 107 45 222

Expected frequencies (fe)


technical mechanical material Total
Morning 21.44 32.77 13.78 68
Afternoon 27.12 41.45 17.43 86
Night 21.44 32.77 13.78 68
Total 70 107 45 222

Chi-Squared components
technical mechanical material
Morning 1.9351 2.5967 0.5622
Afternoon 0.0460 0.0508 0.3782
Night 2.6646 1.8443 0.0034
χ²-stat = 10.0812

(c) Conclusion
Since χ²-stat = 10.0812 > χ²-crit = 9.488, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
the nature of tyre defects produced is related to the shift on which the defects occur.
The nature of the relationship is described in (a) above.
(d) Let πi = proportion of defective tyres caused by mechanical factors per shift i .

H0: π1 = π2 = π3
H1: At least one πi is different (i = 1,2,3)

The hypothesis test procedure is identical to (b) above.


The sample proportions being compared are: Morning Afternoon Night
0.618 0.465 0.368

Conclusion
Since H0 is rejected in favour of H1 at the 5% significance level, it can be concluded that there
is at least one shift that has a different proportion of defective tyres due to mechanical factors.
Based on the row percentages table in (a) above, it is clear that the morning shift
produces a proportionally larger percentage of defective tyres due to mechanical
factors than the afternoon or night shifts.
Exercise 10.17 File: X10.17 - flight delays.xlsx

(a) Histogram
Histogram of Flight Delay Times
Delays(min) Count
<5 0
5 - 7.5 8 35
7.5 - 10 28 28 29
30
10-12.5 29
25
12.5-15 13

No. of flights
15-17.5 2 20
Total 80 13
15

10 8

5 2
0
0
<5 5 - 7.5 7.5 - 10 10-12.5 12.5-15 15-17.5

Delay Intervals (minutes)

Interpretation Flight delay times (in minutes) appear to be normally distributed.

(b) Descriptive Statistics flight delays


Mean 10.324
Standard Error 0.261
Median 10.3
Mode 11
Standard Deviation 2.333
Sample Variance 5.444
Kurtosis -0.550
Skewness 0.084
Range 10.3
Minimum 5.3
Maximum 15.6
Sum 825.9
Count 80

Interpretation The low skewness coefficient (0.084) indicates approximate normality.

(c) (i) Goodness-of-fit test for Normality


H0: Flight time delays (in minutes) follows a normal distribution with μ = 10.324 min and σ = 2.333 min.
H1: Flight time delays (in minutes) do not follow a normal distribution with μ = 10.324 min and σ = 2.333 min.

Region of Rejection Use α = 0.01 with degrees of freedom = (7-2-1) = 4


χ²-crit = χ²(0.01)(4) = 13.277
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 13.277

Expected frequencies using x ≡ N(10.324; 2.333) from Z-table


Delay Intervals fo Normal probability intervals (x and z) Probability fe χ²-stat
-∞ < x < 5 0 P(x < 5) P(z < -2.282) 0.01130 0.90 0.9040
5 < x < 7.5 8 P(5 < x < 7.5) P(-2.282 < z < -1.21) 0.10180 8.14 0.0025
7.5 < x < 10 28 P(7.5 < x < 10) P(-1.21 < z < -0.139) 0.33120 26.50 0.0854
10 < x < 12.5 29 P(10 < x < 12.5) P(-0.139 < z < 0.933) 0.37950 30.36 0.0609
12.5 < x < 15 13 P(12.5 < x < 15) P(0.933 < z < 2.004) 0.15400 12.32 0.0375
15 < x < 17.5 2 P(15 < x < 17.5) P(2.004 < z < 3.076) 0.02117 1.69 0.0554
17.5 < x < +∞ 0 P(x > 17.5) P(z > 3.076) 0.00103 0.08 0.0824
80 1 80
χ²-stat = 1.2282
Conclusion
Since χ²-stat = 1.2282 << χ²-crit = 13.277, there is insufficient sample evidence
at the 1% significance level to reject H 0 in favour of H1. Therefore conclude that
flight delay times (in minutes) follows a Normal distribution with μ = 10.324 min and σ = 2.333 min.
(c) (ii) NORMDIST() From Z-table
P(x < 5) =NORMDIST(5,10.324,2.333,1) 0.01124 0.01130
P(5 < x < 7,5) =NORMDIST(7.5,10.324,2.333,1) - NORMDIST(5,10.324,2.333,1) 0.10181 0.10180
P(7,5 < x < 10) =NORMDIST(10,10.324,2.333,1) - NORMDIST(7.5,10.324,2.333,1) 0.33172 0.33120
P(10 < x < 12,5) =NORMDIST(12.5,10.324,2.333,1) - NORMDIST(10,10.324,2.333,1) 0.37974 0.37950
P(12,5 < x < 15) =NORMDIST(15,10.324,2.333,1) - NORMDIST(12.5,10.324,2.333,1) 0.15297 0.15400
P(15 < x < 17,5) =NORMDIST(17.5,10.324,2.333,1) - NORMDIST(15,10.324,2.333,1) 0.02147 0.02117
P(x > 17,5) =1-NORMDIST(17.5,10.324,2.333,1) 0.00105 0.00103
1 1
Exercise 10.18 File: newspaper sections.xlsx

(a) Two-way Pivot table of Gender by Newspaper Section Most Preferred to Read.

Section
Gender Data Sport Social Business Grand Total
Female Count 14 28 18 60
Row % 23.3% 46.7% 30.0% 100%
Male Count 41 35 44 120
Row % 34.2% 29.2% 36.7% 100%
Total Count 55 63 62 180
Total Row % 30.6% 35.0% 34.4% 100%

Interpretation (by inspection)


Females tend to read the Social section most, with the Sports section read the least.
Males, alternatively, are marginally more interested in the Sport and Business sections.
These observational conclusions are, however, not overwhelmingly conclusive.

(b)
Stacked Bar Chart
Newspaper Section Read by Gender

120%

100%
30.0% 36.7%
80%
Business
60% Social
46.7% 29.2% Sport
40%

20% 34.2%
23.3%
0%
Female Male

(c) Formulate as a test for independence of association between gender and section read .

H0: There is no association between gender and the newspaper section most preferred.
H1 : There is an association between gender and the newspaper section most preferred.

Region of Rejection Use α = 0.10 with degrees of freedom = (2-1)(3-1) = 2


χ²-crit = χ²(0.10)(2) = 4.605
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 4.605
χ²-stat Observed frequencies (fo)
Sport Social Business Total
female 14 28 18 60
male 41 35 44 120
Total 55 63 62 180

Expected frequencies (fe)


Sport Social Business Total
female 18.3 21.0 20.7 60
male 36.7 42.0 41.3 120
Total 55 63 62 180

Chi-Squared components
Sport Social Business
female 1.0242 2.3333 0.3441
male 0.5121 1.1667 0.1720
χ²-stat = 5.5525

Conclusion
Since χ²-stat = 5.5525 > χ²-crit = 4.605 (marginal), there is sufficient sample evidence
at the 10% significance level to reject H0 in favour of H1. Therefore conclude that
gender and the newspaper section most preferred are statistically associated.
The nature of the relationship is described in (a) above.

(d) Let πi = proportion of females who most prefer each newspaper section i .

H0: π1 = π2 = π3
H1: At least one πi is different (i = 1,2,3)

The hypothesis test procedure is identical to (c) above.


The sample proportions being compared are:
Sport Social Business
0.255 0.444 0.290

Conclusion
Since H0 is rejected in favour of H1 at the 10% significance level, it can be concluded that
there is at least one newspaper section that females prefer differently to the other sections.
Based on the row percentages table in (a) above, it is clear that
females tend to read the Social section most, with the Sports section read the least.
Exercise 10.19 File: X10.19 - vehicle financing.xlsx

(a) One-way Pivot table of Car Loan Sizes

Count of Loan size Count of Loan size


Loan size Total Loan size Total
Under 100 18 Under 100 6%
100 - <150 58 100 - <150 19%
150 - <200 110 150 - <200 37%
200 - <250 70 200 - <250 23%
Above 250 44 Above 250 15%
Grand Total 300 Grand Total 100%

Interpretation (by inspection)


The most popular car loan size was between R150 000 and R200 000 (37% of all
applications) followed by car loan sizes of between R200 000 and R250 000 (23%).
Only 6% of loan applications were for amounts below R100 000.

(b)
Bar Chart of Car Loan Applications
0.4

0.35

0.3

0.25

0.2
0.37
0.15
0.23
0.1 0.19
0.15
0.05
0.06
0
Under 100 100 - <150 150 - <200 200 - <250 Above 250
Size of Loan Applications (R 1000)

(c) Test for Goodness-of-Fit for an empirical distribution.

H0: There is no change in the size of car loan applications from four years ago.
H1: There is a significant shift in the size of car loan applications from four years ago.

(d) Region of Rejection Use α = 0.05 with degrees of freedom = (5-1) = 4


χ²-crit = χ²(0.05)(4) = 9.488
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 9.488

χ²-stat Payment Methods


Loan Size fo % fe fe χ²-stat
< R100 18 10 30 4.800
R100 - R150 58 20 60 0.067
R150 - R200 110 40 120 0.833
R200 - R250 70 20 60 1.667
> R250 44 10 30 6.533
Total 300 100 300 13.900
Conclusion
Since χ²-stat = 13.9 > χ²-crit = 9.488, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
there has been a significant shift in the size of car loan applications from 4 years ago.
The shift has been towards larger car loan applications.
Exercise 10.20 File: X10.20 - milk products.xlsx

(a) Two-way Pivot table of milk type purchased and health-concious status of consumers

Question 2
Question 1 Data Yes No Grand Total
Fat-free Count 20 5 25
Percent 43.5% 16.7% 32.9%
Low fat Count 15 10 25
Percent 32.6% 33.3% 32.9%
Full cream Count 11 15 26
Percent 23.9% 50.0% 34.2%
Total Count 46 30 76
Total Percent 100% 100% 100%

Interpretation (by inspection)


The sample evidence points strongly towards health-concious consumers purchasing more
fat-free dairy products, while non-health-concious consumers tend to purchase more
full cream dairy products.

(b)
Stacked Bar Chart
Milk Type Purchased and Health-Concious Status

80%
70%
% of Consumers

60%
50% 16.7% 33.3% 50.0% No
40%
Yes
30%
20% 43.5%
32.6%
10% 23.9%
0%
Fat-free Low fat Full cream

Milk Categories

(c) Formulate as a test for independence of association between milk type purchased and
health-concious status of consumers.

H0: There is no association between milk type purchased and health-concious status.
H1: There is an association between milk type purchased and health-concious status.

Region of Rejection (Use α = 0.05 with degrees of freedom = (3-1)(2-1) = 2)


χ²-crit = χ²(0.05)(2) = 5.991
Decision rule
Reject H0 in favour of H1 if χ²-stat ≥ 5.991
χ²-stat Observed frequencies (fo)
Yes No Total
fat-free 20 5 25
low fat 15 10 25
full cream 11 15 26
Total 46 30 76

Expected frequencies (fe)


Yes No Total
fat-free 15.13 9.87 25
low fat 15.13 9.87 25
full cream 15.74 10.26 26
Total 46 30 76

Chi-Squared components
Yes No
fat-free 1.566 2.402
low fat 0.001 0.002
full cream 1.426 2.186
χ²-stat = 7.583

Conclusion
Since χ²-stat = 7.583 > χ²-crit = 5.991, there is sufficient sample evidence
at the 5% significance level to reject H0 in favour of H1. Therefore conclude that
there is a significant statistical association between the health-concious status
of a consumer and their preference for certain type of dairy milk products.
The nature of the relationship is described in (a) above.

(d) Let πi = proportion of health-concious consumers who prefer each milk type i .

H0: π1 = π2 = π3
H1: At least one πi is different (i = 1,2,3)

The hypothesis test procedure is identical to (c) above.


The sample proportions being compared are: fat-free low-fat full cream
0.800 0.600 0.423

Conclusion
Since H0 is rejected in favour of H1 at the 5% significance level, it can be concluded that
there is at least one milk type that is purchased by a different proportion of health-concious
consumers. Based on the row percentages table in (a) above, it is clear that health-concious
consumers tend to purchase more fat-free dairy products than full cream products.
CHAPTER 11

ANALYSIS OF VARIANCE

COMPARING MEANS ACROSS MULTIPLE POPULATIONS

Exercise 11.1 The purpose of one-factor Anova is to test for equality of means across
multiple (more than two) populations.

Exercise 11.2 Example: Compare the output performance of five identical machines.

Exercise 11.3 Variation between groups measures how similar or how different
(i.e. how close or how far apart) the sample means are from each other.
It is a measure of the level of influence of the treatment factor on the
response measure. Any differences can be attributed to (or explained by)
the influence of the treatment factor on the numeric response measure.

Exercise 11.4 SST = 25.5


Numerator degrees of freedom = (4 - 1) = 3
SSE = 204.6 - 25.5 = 179.1
Denominator degrees of freedom = (4x10 - 4) = 36

MST = 25.5 / 3 = 8.5


MSE = 179.1 / 36 = 4.975

F-stat = 8.5 / 4.975 = 1.7085

Exercise 11.5 F-crit = F(0.05, 3, 36) = 2.866 In Excel, use =FINV(0.05,3,36)

Exercise 11.6 H 0 : μ 1 = μ2 = μ3 = μ4
H1: At least one μi differs

Decision Rule: Do not reject H0 if F-stat ≤ F-crit .


Since F-stat = 1.7085 < F-crit = 2.866, do not reject H0.
Conclusion: All population means are equal.
Exercise 11.7 File: X11.7 - car fuel efficiency.xlsx

(a) Sample Average Fuel Consumption (l/100km)

Peugot VW Ford
Σxi 32.4 29.3 34.4
ni 5 4 5
x(bar)i 6.48 7.325 6.88
Grand mean = 96.1/14 = 6.864

Bar Chart of Fuel Efficiency Means

10
9
8 7.325
6.88
7 6.48
litres / 100 km

6
5
4
3
2
1
0
Peugot VW Ford
Car Types 6.48 7.325 6.88

(b) One-Factor Anova Factor = Car Type ( 1 = Peugot; 2 = VW; 3 = Ford)


Response measure = Fuel Consumption (l/100km)

H0: μ1 = μ2 = μ3
H1: At least one μi differs (i = 1, 2, 3)

Region of Rejection (Use α = 0.05 with df1 = 3-1 = 2 and df2 = 14-3 = 11)
F-crit = F(0.05)(2,11) = 3.98
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 3.98

F-stat SSW = (7-6.48)²+(6.3-6.48)²+(6-6.48)²+(6.4-6.48)²+(6.7-6.48)²+


(6.8-7.325)²+(7.4-7.325)²+(7.9-7.325)²+(7.2-7.325)²+
(7.6-6.88)²+(6.8-6.88)²+(6-4.88)²+(7-6.88)²+(6.6-6.88)²
= 2.0635
SSB = (6.48-6.864)²+(7.325-6.864)²+(6.88-6.864)²
= 1.5886
SST = 2.0635 + 1.5886 = 3.6521

Anova Table
Source of
SS df MS F-stat
Variation
Between 1.5886 2 0.79432 4.234
Within 2.0635 11 0.18759
Total 3.6521 13
Conclusion
Since F-stat = 4.234 > F-crit = 3.98, there is sufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is at
least one motor vehicle type with a different average fuel consumption to the rest.

By inspection, it would appear that VW has an average fuel consumption that is significantly
different (higher, and hence least fuel efficient) from Peugot and Ford.

(c) New Region of Rejection (Use α = 0.01 with df1 = 3-1 = 2 and df2 = 14-3 = 11)
F-crit = F(0.01)(2,11) = 7.21
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 7.21

New decision Do not reject H0 at the 1% significance level,


since F-stat = 4.234 < F-crit = 7.21

New conclusion
There is no statistical evidence, at the 1% signficance level, to conclude that average
fuel consumption differs across motor vehicle types.

Note: The sample evidence must be more convincing (i.e. larger differences between
sample means) before one is prepared to reject the null hypothesis in favour of the
alternative hypothesis.
A level of significance of 1% indicates that the sample evidence is not strong
enough (meaningful differences) yet to reject the null hypothesis of equal means.
Exercise 11.8 File: X11.8 - package design.xlsx

Assumption 1 Equal population variances.


Assumption 2 A normally distribution population for the response variable.

(a) One-Factor Anova Factor = Package designs (1 = A; 2 = B; 3 = C)


Response measure = Carton sales (units)

Let μi = population mean sales of a breakfast cereal packaged in design shape i


H0: μ1 = μ2 = μ3
H1: At least one μi differs (i = 1, 2, 3)

Region of Rejection (Use α = 0.05 with df1 = 3-1 = 2 and df2 = 21-3 = 18)
F-crit = F(0.05)(2,18) = 3.55
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 3.55

F-stat (Refer to formulae in Chapter 11)

Sample evidence
Design A Design B Design C
Sample means 35.75 32.86 34.17
Grand mean 721/21 = 34.33

SSW = (35-35.75)²+(37-35.75)²+(39-35.75)²+(36-35.75)²+(30-35.75)²+(39-35.75)²+(36-35.75)²+(34-35.75)²+
(35-32.86)²+(34-32.86)²+(30-32.86)²+(31-32.86)²+(34-32.86)²+(32-32.86)²+(34-32.86)²+
(38-34.17)²+(34-34.17)²+(32-34.17)²+(34-34.17)²+(34-34.17)²+(33-34.17)²
= 101.1905
SSB = (35.75-34.33)²+(32.86-34.33)²+(34.17-34.33)²
= 31.4762
SST = 101.1905+31.4762 = 132.6667

Anova Table
Source of
SS df MS F-stat
Variation
Between 31.4762 2 15.7381 2.7995
Within 101.1905 18 5.6217
Total 132.6667 20

Conclusion
Since F-stat = 2.7995 < F-crit = 3.55, there is insufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is
no difference in the mean volume of sales across the 3 package designs.

(b) Recommendation
There is no strong statistical evidence to conclude that sales volumes differ across the
three package designs. All are likely to generate the same average sales. Therefore the cereal
producer can choose any of the three package designs for their new muesli cereal.
Exercise 11.9 File: X11.9 - bank service.xlsx

(a) One-Factor Anova Factor = Bank (1 = X; 2 = Y; 3 = Z)


Response measure = Rating score (1 to 10)

Let μi = population mean service rating score for bank i


H0: μ 1 = μ2 = μ3
H1: At least one μi differs (i = 1, 2, 3)

Region of Rejection (Use α = 0.10 with df1 = 3-1 = 2 and df2 = 27-3 = 24)
F-crit = F(0.10)(2,24) = 2.538 =FINV(0.1,2,24)
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 2.538

F-stat (Refer to the formulae in Chapter 11)

Sample evidence
Bank X Bank Y Bank Z
Sample means 6.875 5.778 6.2
Grand mean 169/27 = 6.259

SSW = (8-6.875)²+(6-6.875)²+...+(5-5.778)²+(6-5.778)²+…+(8-6.2)²+(7-6.2)²+…+(6-6.2)²
= 24.0306
SSB = (6.875-6.259)²+(5.778-6.259)²+(6.2-6.259)²
= 5.1546
SST = 24.0306 + 5.1546 = 29.1852

Anova Table
Source of
SS df MS F-stat
Variation
Between 5.1546 2 2.5773 2.574
Within 24.0306 24 1.0013
Total 29.1852 26

Conclusion
Since F-stat = 2.574 > F-crit = 2.538, there is sufficient sample evidence at the
10% significance level to reject H0 in favour of H1. Therefore conclude that there is
at least one bank that has a different mean service rating score to the other banks.

By inspection, it would appear that Bank X has a significantly higher mean


service rating score than the other two banks.

(b) New Region of Rejection (Use α = 0.05 with df1 = 3-1 = 2 and df2 = 27-3 = 24)
F-crit = F(0.05)(2,24) = 3.40
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 3.40

New decision Do not reject H0 at the 5% significance level,


since F-stat = 2.574 < F-crit = 3.40

New conclusion
There is no statistical evidence, at the 5% signficance level, to conclude that the
mean service ratings is different across the three banks.
The three banks are perceived similarly by customers in terms of their service levels.

Note: The reason for the change in conclusion between (a) and (b) is that the
statistical evidence is only weak (i.e. small differences in sample means) at
the 10% significance level, while it the 5% significance level, it is not seen as
strong enough (i.e. differences are not large enough) to reject the null hypothesis.
Exercise 11.10 File: X11.10 - shelf height.xlsx

One-Factor Anova Factor = Shelf Height (1 = Bottom; 2 = Waist; 3 = Shoulder; 4 = Top)


Response measure = Sales volume (units sold)

Let μi = population mean sales of a drinking chocolate product displayed at shelf height i
H0: μ1 = μ2 = μ3 = μ4
H1: At least one μi differs (i = 1, 2, 3, 4)

Region of Rejection (Use α = 0.05 with df1 = 4-1 = 3 and df2 = 30-4 = 26)
F-crit = F(0.05)(3,26) = 2.990
Decision rule
Reject H0 in favour of H1 if F-stat ≥ 2,990

F-stat (Refer to the formulae in Chapter 11)

Sample evidence
Bottom Waist Shoulder Top
Sample means 76.143 81.375 82.444 74.833
Grand mean 2375/30 = 79.167

SSW = (78-76.143)²+….+(78-81.375)²+….+(83-82.444)²+….+(69-74.833)²+…+(75-74.833)²
= 727.788
SSB = (76.143-79.167)²+(81.375-79.167)²+(82.444-79.167)²+(74.833-79.167)²
= 312.379
SST = 727.788 + 312.379 = 1040.167

Anova Table
Source of
SS df MS F-stat
Variation
Between 312.379 3 104.126 3.720
Within 727.788 26 27.992
Total 1040.167 29

Conclusion
Since F-stat = 3.72 > F-crit = 2.99, there is sufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is
at least one shelf height that generates a different mean level of sales to the other shelves.

By inspection, it would appear that shoulder and waist high shelves generate higher average
sales of the drinking chocolate product than bottom or top shelves.
Exercise 11.11 File: X11.11 - machine evaluation.xlsx

(a) One-Factor Anova Factor = Labelling Machine (1 = A; 2 = B; 3 = C)


Response measure = Processing time (in minutes)

Let μi = population mean processing time for shaping and labelling machine i
H0: μ1 = μ2 = μ3
H1: At least one μi differs (i = 1, 2, 3)

Anova: Single Factor

SUMMARY
Groups Count Sum Average Variance
Machine A 5 59 11.8 3.7
Machine B 5 56 11.2 5.7
Machine C 5 72 14.4 3.8

ANOVA
Source of Variation SS df MS F-stat P-value F crit
Between Groups 28.933 2 14.467 3.288 0.0727 2.8068
Within Groups 52.8 12 4.4

Total 81.733 14

Conclusion
Since F-stat = 3.288 > F-crit = 2.8068, there is sufficient sample evidence at the
10% significance level to reject H0 in favour of H1. Therefore conclude that there is at least
machine that has a different mean processing time.

By inspection, machine C has a significantly longer mean processing time than either
machines A and B. Machine C must not be considered for purchase.

(b) Two-Sample Test of Mean Processing Times between Machines A (1) and B (2)
H0: μ1 = μ2 (two tailed test)
H1: μ1 ≠ μ2

t-Test: Two-Sample Assuming Equal Variances

Machine A Machine B
Mean 11.8 11.2
Variance 3.7 5.7
Observations 5 5
Pooled Variance 4.7
Hypothesized Mean Difference 0
df 8
t Stat 0.4376
P(T<=t) one-tail 0.3366
t Critical one-tail 1.3968
P(T<=t) two-tail 0.6733 t-crit (0.10, 8) =
t Critical two-tail 1.8595 =TINV(0.1,8) 1.8595
Conclusion
Since t-stat = 0.4376 lies with the region of non-rejection of H 0 (i.e. within ± 1.8595),
the sample evidence is not strong enough to reject H 0 in favour of H1.
The population mean processing times between the two machines A and B are
likely to be identical. Thus the company can purchase either machine A or machine B.

(c) Recommendation
Based on the statistical evidence in (a) and (b), the company can purchase either
machine A or machine B - they are likely to be equally efficient in operation.
Exercise 11.12 File: X11.12 - earnings yield.xlsx

(a) One-Factor Anova Factor = Economic Sector


(1 = Financial; 2 = Retail; 3 = Industrial; 4 = Mining)
Response measure = Earnings yield (%)

Let μi = population mean earnings yield per economic sector i


H0: μ 1 = μ 2 = μ 3 = μ4
H1: At least one μi differs (i = 1, 2, 3, 4)

(b) Anova: Single Factor

SUMMARY
Groups Count Sum Average Variance
Industrial 20 96.1 4.81 1.091
Retail 20 94.4 4.72 1.492
Financial 20 113.2 5.66 1.588
Mining 20 111 5.55 2.067

ANOVA
Source of Variation SS df MS F-stat P-value F crit
Between Groups 14.3894 3 4.7965 3.0757 0.0326 2.7249
Within Groups 118.5195 76 1.5595

Total 132.9089 79

Conclusion
Since F-stat = 3.076 > F-crit = 2.725, there is sufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is
at least one sector with a different mean earnings yield relative to the other sectors.

(c) Interpretation
By inspection, the Industrial and Retail sectors each appear to have significantly lower
mean earnings yields relative to the Financial and Mining sectors' mean earnings yields.
Exercise 11.13 File: X11.13 - advertising strategy.xlsx

(a) One-Factor Anova


Factor = Advertising strategy
(1 = Sophisticated; 2 = Athletic; 3 = Trendy)
Response measure = Sales (units of cans)

Let μi = population mean level of sales achieved under each advertising strategy i
H0: μ1 = μ2 = μ3
H1: At least one μi differs (i = 1, 2, 3)

(b) Anova: Single Factor

SUMMARY
Groups Count Sum Average Variance
Sophisticated 20 8380 419 5075.579
Athletic 20 7241 362.05 6201.313
Trendy 20 8030 401.5 3781.737

ANOVA
Source of Variation SS df MS F-stat P-value F-crit
Between Groups 34039.03 2 17019.52 3.391 0.0406 3.159
Within Groups 286113.95 57 5019.54

Total 320152.98 59

Conclusion
Since F-stat = 3.391 > F-crit = 3.159, there is sufficient sample evidence at the 5% level of
significance to reject H0 in favour of H1. Therefore conclude that there is at least one
advertising strategy that results in a different mean level of deodorant sales relative
to the other strategies.

(c) Interpretation and recommendation


By inspection, the Athletic advertising strategy resulted in the lowest mean sales of all
three advertising strategies. On average, the Sophisticated and Trendy strategies
appear to be equally effective (the difference in sample means does not appear significant).

Recommendation:
Either the Trendy or the Sophisticated strategy can be adopted.
(d) Two-Sample Test of Means between the Sophisticated (1) and Trendy (2) Strategies
H0: μ1 = μ2 (two-tailed test)
H1: μ1 ≠ μ2

t-Test: Two-Sample Assuming Equal Variances

Sophisticated Trendy
Mean 419 401.5
Variance 5075.579 3781.737
Observations 20 20
Pooled Variance 4428.6579
Hypothesized Mean Difference 0
df 38
t Stat 0.8316
P(T<=t) one-tail 0.2054
t Critical one-tail 1.6860
P(T<=t) two-tail 0.4108 t-crit (0.05,38) =
t Critical two-tail 2.0244 =TINV(0.05,38) 2.0244

Conclusion
Since t-stat = 0.8316 lies within the region of non-rejection of H0 (i.e. ± 2.0244), there is
insufficient sample evidence to reject H0 in favour of H1.
Therefore the population mean sales from each of the two strategies is likely to be identical.
The two strategies are therefore equally effective and either can be adopted by the company.

(e) Recommendation
The Athletic strategy can be discarded. It appears the least effective.
The remaining two strategies (Sophisticated and Trendy) are equally effective and
therefore either can be adopted by the company to promote its new ladies deodorant.
Exercise 11.14 File: X11.14 - leverage ratio.xlsx

(a) Descriptive Statistics


Technology Construction Banking Manufacturing

Mean 73.83 78.07 69.73 76.37


Standard Error 1.697 1.912 2.527 2.347
Median 72.5 78.5 68 79.5
Mode 60 80 68 81
Standard Deviation 9.296 10.471 13.841 12.853
Sample Variance 86.420 109.651 191.582 165.206
Kurtosis -1.309 -0.546 -0.453 0.495
Skewness 0.010 0.051 0.103 -0.857
Range 28 40 58 50
Minimum 60 58 41 44
Maximum 88 98 99 94
Sum 2215 2342 2092 2291
Count 30 30 30 30
Confidence Level(95.0%) 3.471 3.910 5.168 4.799

Interpretation
The mean leverage ratio is lowest for the banking sector and highest for the
construction and the manufacturing sectors.
These differences do appear to be significant.

(b) One-Factor Anova Factor = Economic Sector


(1 = Technology; 2 = Construction; 3 = Banking; 4 = Manufacturing)
Response measure = Leverage ratio

Let μi = population mean leverage ratio per economic sector i


H0: μ1 = μ2 = μ3 = μ 4
H1: At least one μi differs (i = 1, 2, 3, 4)

Anova: Single Factor

SUMMARY
Groups Count Sum Average Variance
Technology 30 2215 73.83 86.42
Construction 30 2342 78.07 109.65
Banking 30 2092 69.73 191.58
Manufacturing 30 2291 76.37 165.21

ANOVA
Source of Variation SS df MS F-stat P-value F crit
Between Groups 1181.13 3 393.711 2.849 0.0406 2.683
Within Groups 16032.87 116 138.214

Total 17214 119


Conclusion
Since F-stat = 2.849 > F-crit = 2.683 (marginally), there is sufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is
at least one sector with a different mean leverage ratio relative to the other sectors.

By inspection, the banking sector has the lowest mean leverage ratio, while construction
and manufacturing appear to have similarly high mean leverage ratios.

Recommendation The investor is advised to consider either the banking sector (with the lowest
mean leverage ratio) or the technology sector (with a marginally higher mean
leverage ratio). (This difference may not be statistically significant).

(c) Two-Sample Test of Means - Leverage ratios between Technology (1) and Banking (2) sectors.
H0: μ1 = μ2 (two-tailed test)
H1: μ1 ≠ μ2

t-Test: Two-Sample Assuming Equal Variances

Technology Banking
Mean 73.833 69.733
Variance 86.420 191.582
Observations 30 30
Pooled Variance 139.001
Hypothesized Mean Difference 0
df 58
t Stat 1.347
P(T<=t) one-tail 0.092
t Critical one-tail 1.672
P(T<=t) two-tail 0.183 t-crit (0.05, 58) =
t Critical two-tail 2.002 =TINV(0.05,58) 2.002

Conclusion
Since t-stat = 1.347 lies within the region of non-rejection of H 0 (i.e. within ±2.002), the sample evidence is
not strong enough to reject H 0 in favour of H1. The population mean leverage ratios between the
Technology and the the Banking sector are therefore likely to be equal.

(d) Recommendation
Based on the statistical evidence in (c) and (d), an investor is advised to consider either
the Technology sector or the Banking sector. Their mean leverage ratios are likely to be equal.
Since both sectors offer an investor the same lower risk, either or both can be chosen for investment.
Exercise 11.15 File: X11.15 - training methods.xlsx

(a) Descriptive Statistics


On-the-Job Lecture Role Play Audio-Visual
Mean 9 8.75 9.2 8.85
Standard Error 0.082 0.112 0.094 0.114
Median 9 8.65 9.2 8.9
Mode 9 8.5 9.6 8.9
Standard Deviation 0.327 0.420 0.325 0.426
Sample Variance 0.107 0.177 0.105 0.181
Kurtosis 0.109 0.339 -1.420 3.419
Skewness 0.472 0.759 0.229 -1.522
Range 1.2 1.5 0.900 1.7
Minimum 8.5 8.2 8.8 7.7
Maximum 9.7 9.7 9.7 9.4
Sum 144 122.5 110.4 123.9
Count 16 14 12 14
Confidence Level(95.0%) 0.174 0.243 0.206 0.246

Interpretation
The differences in mean performance scores appear marginal across the four
different training methods, with on-the-job training and role-play having the
highest average scores.

(b) One-Factor Anova Factor = Training Method


(1 = On-the-Job; 2 = Lecture; 3 = Role Play; 4 = Audio-Visual)
Response measure = Performance score (1 - 10)

Let μi = population mean performance score per training method i


H0: μ1 = μ2 = μ3 = μ4
H1: At least one μi differs (i = 1, 2, 3, 4)

Anova: Single Factor

SUMMARY
Groups Count Sum Average Variance
On-the-Job 16 144 9 0.107
Lecture 14 122.5 8.75 0.177
Role Play 12 110.4 9.2 0.105
Audio-Visual 14 123.9 8.85 0.181

ANOVA
Source of Variation SS df MS F-stat P-value F-crit
Between Groups 1.487 3 0.4957 3.479 0.0223 2.7826
Within Groups 7.41 52 0.1425

Total 8.897 55
Conclusion
Since F-stat = 3.479 > F-crit = 2.7826, there is sufficient sample evidence at the
5% significance level to reject H0 in favour of H1. Therefore conclude that there is at least one
training method with a different mean performance score relative to the other training methods.

By inspection, the lecture and audio-visual are the least effective (lower mean scores),
while on-the-job and the role-play methods are more effective (with higher mean scores).

Recommendation
The training manager is advised to consider either on-the-job training or use
role play methods (The difference does not appear to be statistically significant).

(c) Two-Sample Test of Means


Performance scores between On-the-Job (1) and Role Play (2) methods.
H0: μ1 = μ2 (two-tailed test)
H1: μ1 ≠ μ2

t-Test: Two-Sample Assuming Equal Variances

On-the-Job Role Play


Mean 9 9.2
Variance 0.10666667 0.105454545
Observations 16 12
Pooled Variance 0.10615385
Hypothesized Mean Difference 0
df 26
t Stat -1.6074361
P(T<=t) one-tail 0.06001825
t Critical one-tail 1.7056179
P(T<=t) two-tail 0.1200365 t-crit (0.05,26) =
t Critical two-tail 2.05552942 =TINV(0.05,26) 2.0555

Conclusion
Since t-stat = -1.6074 lies within the region of non-rejection of H 0, (i.e. within ±2.0555) the sample
evidence is not strong enough to reject H0 in favour of H1. The population mean performance scores
between the two training methods is likely to be the same.

(d) Recommendation
Based on the statistical evidence in (b) and (c), the HR manager is advised to select either
on-the-job training or role play methods of training.
Both are likely to produce similar high mean performance scores.
Exercise 11.16 In one-factor ANOVA, only one categorical factor is used to explain
possible differences between the observed sample means.
In two-factor ANOVA, two categorical factors are used to explain
possible differences between the observed sample means.

Exercise 11.17 Response variable (y) numeric, ratio-scaled


Factor 1 (x1) categorical (nominal / ordinal-scaled)
Factor 2 (x2) categorical (nominal / ordinal-scaled)

Exercise 11.18 Interaction term means that different combinations of the levels from the two factors
can have different influences (effects) on the numeric response variable.

Exercise 11.19 Interaction plot: It visually displays the nature and strength of the interaction
effect of combinations of factor levels on the dependent response variable.
It is constructed from the sample means of the various combinations
of the different factor levels.
Exercise 11.20

(a) - (f) Two-factor ANOVA with Interaction Table


Source of Variation df SS MS F-stat p -value F -crit
Factor A: PC O.S. 2 68 34 6.8 0.0025 5.08
Factor B: Laptop 3 42 14 2.8 0.0499 4.22
Interaction (AxB) 6 132 22 4.4 0.0013 3.20
Error (Residual) 48 240 5
Total 59 482

(g) Factor A is statistically signficant (at α = 0.01) since F-stat (6.8) > F-crit (5.08).
Alternatively, its p-value (0.0025) < α (0.01).
Factor B is not statistically signficant (at α = 0.01) since F-stat (2.8) < F-crit (4.22).
Alternatively, its p-value (0.0499) > α (0.01).
The interaction effect between factor A and factor B is statistically signficant (at α = 0.01)
since F-stat (4.4) > F-crit (3.20).
Alternatively, its p-value (0.0013) < α (0.01).
XS11.21

Exercise 11.21 File: X11.21 - sales ability.xlsx

(a) (i) Factor: Qualifications Factor: Experience


H0: μ(b) = μ(a) = μ(s) H0: μ(under 3) = μ(over 3)
H1: At least one μi differs (I = b, a, s) H1: At least one μi differs (i = under 3, over 3)

Factor: Interaction
H0: No interaction effect
H1: There is an interaction effect

Anova: Two-Factor With Replication (Data Analysis - Excel )

SUMMARY Business Arts Social Science Total


Under 3 years
Count 10 10 10 30
Sum 400.1 376.6 325.9 1102.6
Average 40.01 37.66 32.59 36.7533333
Variance 61.31656 23.07156 14.56766667 40.628092
Over 3 years
Count 10 10 10 30
Sum 446.2 357.6 415.8 1219.6
Average 44.62 35.76 41.58 40.6533333
Variance 28.37956 92.68711 42.13066667 64.626023
Total
Count 20 20 20
Sum 846.3 734.2 741.7
Average 42.315 36.71 37.085
Variance 48.08029 55.78305 48.12555263

ANOVA
Source of Variation SS df MS F-stat p-value F-crit
Sample (Experience) 228.15 1 228.150 5.222 0.0263 4.020
Columns (Qualifications) 392.73 2 196.365 4.494 0.0157 3.168
Interaction (Exper x Qualif) 300.26 2 150.131 3.436 0.0394 3.168
Within 2359.38 54 43.692
Total 3280.52 59

(a) (ii) Factor: Qualifications Since F-stat (4.494) > F-crit (3.168), reject H0 and conclude that qualifications
is a statistically significant factor (α = 0.05)
Business graduates generate significantly higher average sales (42.32)
than graduates with Arts (36.71) or Social Science (37.09) degrees.

(a) (iii) Factor: Experience Since F-stat (5.22) > F-crit (4.02), reject H0 and conclude that experience
is a statistically significant factor (α = 0.05)
Graduates with over 3 years experience generate significantly higher average sales (40.65)
than graduates with less than 3 years experience (36.75).

(a) (iv) Factor: Interaction effect Since F-stat (3.436) > F-crit (3.168), reject H0 and conclude that there is a significant
interaction effect between experience and qualifications on graduates' sales performance (α = 0.05)
Business graduates with over 3 years experience perform the best (44.62) while social science
graduates with with less than 3 years experience perform the worst (32.59).

(b) Summary of Sample Means Table

Business Arts Social Science


Under 3 years 40.01 37.66 32.59
Over 3 years 44.62 35.76 41.58

Page 19
XS11.21

Chart Title
46 Interaction Plot
44
42
40
38
36
34
32
30
Business Arts Social Science

Under 3 years Over 3 years

See (a) (iv) for an interpretation of the interaction plot.

(c) Recommendation to HR manager:


Recruit predominantly business and social science graduates with over 3 years of work experience.
If Arts graduates are employed, they must be given intensive marketing training.

Page 20
Exercise 11.22 File: X11.22 - dropped calls.xlsx

(a) (i) Random variable: % of daily dropped calls per network switch and transmission type
Factor 1 = Switch type (SW1, SW2, SW3, SW4)
Factor 2 = Transmission type (Voice, Data bundles)

Management Question: Is the % of daily dropped calls the same across all network switching
devices and / or transmission types

Factor: Switch type Factor: Transmission type


H0: μ(1) = μ(2) = μ(3) = μ(4) H0: μ(voice) = μ(data)
H1: At least one μi differs (i = 1,2,3,4) H1: At least one μi differs (i = voice, data)

Factor: Interaction
H0: No interaction effect
H1: There is an interaction effect

Anova: Two-Factor With Replication (Data Analysis - Excel )

SUMMARY SW1 SW2 SW3 SW4 Total


Voice
Count 8 8 8 8 32
Sum 7.76 4.35 4.69 6.39 23.19
Average 0.97 0.54375 0.58625 0.79875 0.724688
Variance 0.098657 0.081884 0.086227 0.071155 0.106645
Data
Count 8 8 8 8 32
Sum 5.83 6.82 8.16 7.1 27.91
Average 0.72875 0.8525 1.02 0.8875 0.872188
Variance 0.047984 0.078707 0.057171 0.067279 0.067818
Total
Count 16 16 16 16
Sum 13.59 11.17 12.85 13.49
Average 0.849375 0.698125 0.803125 0.843125
Variance 0.083953 0.100363 0.11709 0.066703

ANOVA
Source of Variation SS df MS F-stat P-value F-crit
Sample (Transmission) 0.3481 1 0.3481 4.727498 0.033926 4.012973
Columns (Switch) 0.234819 3 0.078273 1.063014 0.372175 2.769431
Interaction (Trans x Switch) 1.050075 3 0.350025 4.753641 0.005055 2.769431
Within 4.12345 56 0.073633
Total 5.756444 63

(a) (ii) Factor: Switching device Since F-stat (1.063) < F-crit (2.769), do not reject H0 and conclude that switch devices are not statistically
significant (α = 0.05). Hence all switch devices are likely to have the same average dropped call rate.

(a) (iii) Factor: Transmission Since F-stat (4.73) > F-crit (4.01), reject H0 and conclude that transmission type
is a statistically significant factor (α = 0.05)
Data transmission lead to a higher average dropped call rate (0.872) than voice transmission (0.725).

(a) (iv) Factor: Interaction effect Since F-stat (4.75) > F-crit (2.769), reject H0 and conclude that there is a significant
interaction effect between switch devices and transmission type. (α = 0.05)
SW2 and SW3, transmitting voice (0.544 and 0.586), are likely to have the lowest average dropped call rate,
while SW1 transmitting voice (0.97) and SW3 transmitting data (1.02), are likely to have the
highest average dropped call rate.
(b) Summary of Sample Means Table

SW1 SW2 SW3 SW4


Voice 0.970 0.544 0.586 0.799
Data 0.729 0.853 1.020 0.888

Chart Title
1.100
1.000 1.020
0.970
0.900 0.888
0.853
0.800 0.799
0.700 0.729

0.600 0.586
0.544
0.500
0.400
SW1 SW2 SW3 SW4

Voice Data

See (a) (iv) for an interpretation of the interaction plot.

(c) Recommendation to the chief engineer:


For voice transmissions, use switching devices 2 and 3;
For data transmission, use only switching device 1.
Investigate the high % dropped calls rates of switching device 3 (for data transmission)
and switching device 4 for both transmission types.
Exercise 11.23 File: X11.23 - rubber wastage.xlsx

(a) Random variable: % rubber wastage per week

Factor: Machine (TAM) Factor: Tyre


H0: μ(1) = μ(2) = μ(3) H0: μ(R) = μ(B)
H1: At least one μi differs (i = 1, 2, 3) H1: At least one μi differs (i = Radial, Bias)

Factor: Interaction
H0: No interaction effect
H1: There is an interaction effect

Anova: Two-Factor With Replication (Data Analysis - Excel )

SUMMARY TAM1 TAM2 TAM3 Total


Radial
Count 5 5 5 15
Sum 19.71 33.52 19.44 72.67
Average 3.942 6.704 3.888 4.844667
Variance 3.96097 6.56833 3.44237 5.844455
Bias
Count 5 5 5 15
Sum 13.47 38.6 43.77 95.84
Average 2.694 7.72 8.754 6.389333
Variance 1.35283 9.83965 8.96443 13.26548
Total
Count 10 10 10
Sum 33.18 72.12 63.21
Average 3.318 7.212 6.321
Variance 2.794329 7.579173 12.09134

ANOVA
Source of Variation SS df MS F-stat p-value F-crit
Sample (Tyres) 17.89496 1 17.89496 3.146037 0.088802 4.259677
Columns (TAMS) 83.25042 2 41.62521 7.317951 0.003301 3.402826
Interaction (Tyre x TAM) 47.77433 2 23.88716 4.1995 0.0273 3.402826
Within 136.5143 24 5.688097
Total 285.434 29

Statistical and Management Conclusions


Factor: Tyres Since F-stat (3.15) < F-crit (4.26), do not reject H0 and conclude that tyre types is not
a statistically significant factor (α = 0.05)
Regardless of TAM used, average rubber wastage is the same across both tyre types produced.
Factor: TAMs Since F-stat (7.318) > F-crit (3.403), reject H0 and conclude that TAM used
is a statistically significant factor (α = 0.05)
Regardless of tyre type produced, average rubber wastage is likely to be lowest on TAM1 (3.318).
compared to TAM2 (7.212) and TAM3 (6.321).
Factor: Interaction effect Since F-stat (4.199) > F-crit (3.403), reject H0 and conclude that there is a significant
interaction effect between tyre type produced and TAM used (α = 0.05)
Average rubber wastage for bias tyres is lowest on TAM1 (2.694) but highest on TAM3 (8.754).
By contrast, average rubber wastage of radial tyres is lowest on TAM3 (3.888) and highest on TAM2 (6.704).

(b) Interaction Plot TAM1 TAM2 TAM3


Radial 3.942 6.704 3.888
Bias 2.694 7.720 8.754

Chart Title
10.000
9.000 8.754
8.000 7.720
7.000 6.704
6.000
5.000
4.000 3.942 3.888
3.000 2.694
2.000
1.000
0.000
TAM1 TAM2 TAM3

Radial Bias

See (a) for an interpretation of the interaction plot.


(c) Recommendation to the production manager:
TAM1 machine is the most efficient in minimising wastage for both tyre types (purchase more of TAM1).
Allocate Radial tyres manufacturing to TAM3 (it has the minimum % wastage of all three machines on Radial tyres).
Investigate the high % wastage of TAM3 for Bias tyres and TAM2 for both tyre types.
Possibly replace TAM2 type machines with TAM1 type machines.
CHAPTER 12

SIMPLE LINEAR REGRESSION AND CORRELATION ANALYSIS

Exercise 12.1 Regression analysis defines the structural relationship between two numeric
variables as a mathematical equation (usually a straight line equation).
Its purpose is to use the equation for estimation / prediction purposes.

Correlation analysis measures the strength of the relationship between the


two numeric variables used in the regression equation.

Exercise 12.2 Dependent variable, y

Exercise 12.3 An independent variable, x , is used as a predictor of the dependent variable.

Exercise 12.4 Scatter plot

Exercise 12.5 Method of Least Squares

Exercise 12.6 Strong inverse linear relationship

Exercise 12.7 H0: ρ = 0 H1: ρ ≠ 0

Degrees of freedom = 18 - 2 = 16
t-crit = t(0.05, 16) = 2.12 (Use =TINV(0.05,16) in Excel )
Decision Rule: Accept H0 if -2.12 ≤ t-stat ≤ +2.12

t-stat = (0.42) * √[(18 - 2)/(1 - 0.422)] = 1.851


Conclusion: Do not reject H0. There is no statistically
significant relationship between x and y .
Exercise 12.8 File: X12.8 - training effectiveness.xlsx

(a) Scatter plot - Training hours versus Productivity

Scatter Plot - Training hours vs Productivity

80
output (units) 70
60
50
40
30
20
10 15 20 25 30 35 40 45

hours of training

Interpretation
There is a strong positive relationship between hours of training
and worker output. The more the training received, the higher the output.

(b) Training (x) Output (y) x² xy y²


20 40 400 800 1600
36 70 1296 2520 4900
20 44 400 880 1936
38 56 1444 2128 3136
40 60 1600 2400 3600
33 48 1089 1584 2304
32 62 1024 1984 3844
28 54 784 1512 2916
40 63 1600 2520 3969
n = 10 24 38 576 912 1444
Σ 311 535 10213 17240 29649

Coefficients b1 = (10*17240-(311*535))/(10*10213-(311²))
1.112
b0 = (535 - 1.112*311)/10
18.917

Thus ŷ = 18.917 + 1.112 x where 20 ≤ x ≤ 40


(c) Correlation coefficient (r) = (17240-(311*535)/10)/√((10213-(311²)/10)*(29649-(535²)/10))
= 0.8072
Coefficient of Determination (r²) = 0.8072² = 0.6516

Variation in training hours can explain 65.16% of the variability in worker output.
This is a high level of explained variation. Hence training input is very beneficial
to worker output and the training programmes should be continued.

(d) x = 25 ŷ =18.917+1.112*25 = 46.717 units


A worker with 25 hours of training can be expected to produce 46.72 units of output, on average.
Exercise 12.9 File: X12.9 - capital utilisation.xlsx

(a) Scatter plot - Earnings Yield versus Inventory Turnover

Scatter Plot
Earnings Yield vs Inventory Turnover

18
16
14
earnings yield

12
10
8
6
4
2
0
0 1 2 3 4 5 6 7 8 9

inventory turnover

Interpretation
There is a strong positive relationship between inventory turnover and earnings yield.
As inventory turnover increases, earnings yields also increases.

(b) Inv t/o (x) E.Y. (y) x² xy y²


3 10 9 30 100
5 12 25 60 144
4 8 16 32 64
7 13 49 91 169
6 15 36 90 225
4 10 16 40 100
8 16 64 128 256
6 13 36 78 169
n =9 5 10 25 50 100
Σ 48 107 276 599 1327

Coefficients b1 = (9*599-(48*107))/(9*276-(48²))
1.4167

b0 = (107-1.4167*48)/9
4.333

Thus ŷ = 4.333 + 1.4167 x where 3 ≤ x ≤ 8

(c) Correlation coefficient (r) = (599-(48*107)/9)/√((276-(48²)/9)*(1327-(107²)/9))


= 0.8552

There is a strong positive linear association between inventory turnover and earnings yield.
Thus the business analyst's view is supported by the strong sample evidence.
(d) Coefficient of Determination (r²) = 0.8552² = 0.7313 i.e. 73.13%

Inventory turnover (capital utilisation) can explain 73.12% of the variability in a company's
earnings yield. This is a high level of explained variation. Hence inventory turnover has been
shown to have a significant direct effect on a company's earnings yield.
Yes, the regression equation can by used with confidence to estimate earnings yield
based on a company's level of inventory turnover.

(e) H0: ρ = 0 Area of Acceptance for H0


H1: ρ ≠ 0
t-crit = t(α=0.05,df = 7) = ± 2.364 2.364624
-2.364 ≤ t-stat ≤ +2.364

t-stat = (0.8552)*√((9-2)/(1-(0.8552²))) = 4.3655

Conclusion
Since t-stat (4.3655) lies outside the area of acceptance for H 0,there is
sufficient sample evidence at the 5% significance level to reject H0 in favour of H1.
Conclude that there is a strong positive association between inventory turnover
and earnings yield.

(f) Expected earnings yield for x = 6 ŷ = 4.333 + 1.4167(6) = 12.833%

A company with an expected inventory turnover of 6 next year can expect to achieve
an earnings yield of 12.833%.
Exercise 12.10 File: X12.10 - loan applications.xlsx

(a) Dependent variable (y) = No. of Loan applications received


Independent variable (x) = interest rate (%)
Interest rate (%) is assumed to influence the number of loan applications received.

(b) Scatter plot

Scatter Plot
Loan applications Received vs Interest Rate (%)

35
no. loan applications received

30

25

20

15

10
4 5 6 7 8 9 10

interest rate (%)

Interpretation
A moderate to strong negative linear relationship between interest rate and
number of loan applications received is observed from the scatter plot.

(c) Correlation coefficient

Int rate % (x) Applications (y) x² y² xy


7 18 49 324 126
6.5 22 42.25 484 143
5.5 30 30.25 900 165
6 24 36 576 144
8 16 64 256 128
8.5 18 72.25 324 153
6 28 36 784 168
6.5 27 42.25 729 175.5
7.5 20 56.25 400 150
8 17 64 289 136
n = 11 6 21 36 441 126
Σ 75.5 241 528.25 5507 1614.5

r= (11*1614.5 - 75.5*241)/√((11*528.25-75.5²)*(11*5507-241²))
= -0.8302
Interpretation
This correlation shows a strong negative (inverse) assocation between
interest rates (%) and the number of loan applications received.
(d) H0: ρ = 0 Area of Acceptance for H0
H1: ρ ≠ 0
t-crit = t(α=0.05,df = 9) = ± 2.262
-2.262 ≤ t-stat ≤ +2.262

t-stat = (-0.8302)*√((11-2)/(1-(-0.8302²))) =
'= -4.4677
Conclusion
Since t-stat (-4,4677) lies outside the area of acceptance for H0,there is
sufficient sample evidence at the 5% significance level to reject H0 in favour of H1.
Conclude that there is a strong negative relationship between interest rates (%)
and the number of loan applications received monthly.

(e) Regression coefficients

b1 = (11*1614.5 - 75.5*241)/(11*528.25-75.5²)
-3.9457

b0 = (241 - (-3.9457)*(75.5))/11
48.991

Thus ŷ = 48.991 - 3.9457 x where 5.5 ≤ x ≤ 8.5

(f) Interpretation of b1 coefficient


For a 1% increase in interest rate, 3.95 fewer loan applications will be received.

(g) x=6 ŷ= 48.991 - 3.9457*(6) = 25.32 applications

If the rate of interest is 6%, the bank can expect to receive 25.32 (say, 25) applications.
Exercise 12.11 File: X12.11 - maintenance costs.xlsx

(a) Dependent variable (y) = Annual maintenance cost (Rand)


Independent variable (x) = Machine age (years)
Machine age is assumed to influence annual maintenance costs.

(b) Scatter plot

Scatter Plot
Annual Maintenance Cost (R) vs Age of Machines

70
annual maintenance cost (R)

60
50
40
30
20
10
0
0 2 4 6 8 10

machine age (years)

Interpretation
A strong direct (positive) linear relationship between the ages of machines and
their annual maintenance costs (in Rands) is observed in the scatter plot.

(c) Correlation coefficient

Age (yrs) Annual Cost (R) x² y² xy


4 45 16 2025 180
3 20 9 400 60
3 38 9 1444 114
8 65 64 4225 520
6 58 36 3364 348
7 50 49 2500 350
1 16 1 256 16
1 22 1 484 22
5 38 25 1444 190
2 26 4 676 52
4 30 16 900 120
n = 12 6 35 36 1225 210
Σ 50 443 266 18943 2182

r= (12*2182 - 50*443)/√((12*266-50²)*(12*18943-443²))
= 0.870028
Interpretation
There is a very strong positive association between the ages of
of machines and their annual maintenance costs (in Rands).
(d) H0: ρ = 0 Area of Acceptance for H0
H1: ρ ≠ 0
t crit = t(α=0.05,df = 10) = ± 2.228
-2.228 ≤ t-stat ≤ +2.228

t-stat = (0.870028)*√((12-2)/(1-0.870028²)) 5.5806

Conclusion
Since t-stat (5.5806) lies outside the area of acceptance for H 0,there is
sufficient sample evidence at the 5% significance level to reject H 0 in favour of H1.
Conclude that there is a strong direct association between age of machines
and the level of annual maintance costs (in Rands).

(e) Regression coefficients

b1 = (12*2182 - 50*443)/(12*266 - 50²)


5.8295
b0 = (443 - 5.8295*50)/12
12.627

Thus ŷ = 12.627 + 5.8295 x where 1 ≤ x ≤ 8

(f) Interpretation of b1 coefficient


If the age of a machine increased by 1 year, annual maintenance costs will rise by R5.8295.
(Alternatively, for every year older, the annual maintenance costs increase by R5.8295)

(g) x=5 ŷ= 12.627 + 5.8295*(5) = R 41.77


For a 5 year old machine, annual maintenance costs are expected to be R41.77.
Exercise 12.12 File: X12.12 - employee performance.xlsx

(a) Scatter Plot

Scatter Plot
Performance Rating vs Aptitude Score

performance rating (0 - 100) 95


90
85
80
75
70
65
60
55
50
2 3 4 5 6 7 8 9 10

aptitude score (1 - 10)

Interpretation
There appears to be a weak to moderate positive association between an employee's
aptitude score and their performance rating.

(b) Correlation coefficient (r)

Aptitude (x) Perf rating (y) x² y² xy


7 82 49 6724 574
6 74 36 5476 444
5 82 25 6724 410
4 68 16 4624 272
5 75 25 5625 375
8 92 64 8464 736
7 86 49 7396 602
8 69 64 4761 552
9 85 81 7225 765
6 76 36 5776 456
4 72 16 5184 288
n = 12 6 64 36 4096 384
Σ 75 925 497 72075 5858

r= (12*5858 - 75*925)/√((12*497-75²)*(12*72075-925²))
0.5194
Interpretation
There is a moderate positive association between the aptitude scores of
employees and their performance scores after one year.
(c) H0: ρ = 0 Area of Acceptance for H0
H1: ρ ≠ 0
t-crit = t(α=0.05,df = 10) = ± 2.228
-2.228 ≤ t-stat ≤ +2.228

t-stat = (0.5194)*√((12-2)/(1-0.5194²)) 1.922

Conclusion
Since t-stat (1.922) lies inside the area of acceptance for H0,there is
not sufficient sample evidence at the 5% significance level to reject H 0 in favour of H1.
Conclude, at a 5% significance level, that there is no statistically significant association
between an employee's aptitude score and their job performance rating score one year later.

(d) Regression coefficients

b1 = (12*5858-75*925)/(12*497-75²) 2.7168

b0 = (925 - 2.7168*75)/12 60.1032

Thus ŷ = 60.1032 + 2.7168 x where 4 ≤ x ≤ 9

Interpretation of b1 coefficient
If an employee's aptitude score increases by one unit, their performance rating
score will increase by 2,7168 points.

(e) Estimation
x =8 ŷ= 60.1032 + 2.7168*(8) = 81.84
For an employee with an aptitude score of 8, they could expect a job performance rating
score of 81.84.

The association between aptitude score and performance rating is not statistically significant.
Therefore, the call centre manager should have low confidence in this
estimated performance rating score.
Exercise 12.13 File: X12.13 - opinion polls.xlsx

(a) Correlation coefficient (r)

Poll (%) (x) Election (%) (y) x² y² xy


42 51 1764 2601 2142
34 31 1156 961 1054
59 56 3481 3136 3304
41 49 1681 2401 2009
53 68 2809 4624 3604
40 35 1600 1225 1400
65 54 4225 2916 3510
48 52 2304 2704 2496
59 54 3481 2916 3186
38 43 1444 1849 1634
n = 11 62 60 3844 3600 3720
Σ 541 553 27789 28933 28059

r= (11*28059 - 541*553)/√((11*27789-541²)*(11*2893-553²))
0.7448
Interpretation
There is a moderate to strong positive association between opinion poll predictions
and the actual election results.

(b) H0: ρ = 0 Area of Acceptance for H0


H1: ρ ≠ 0
t-crit = t(α=0.05,df = 9) = ± 2.262
-2.262 ≤ t-stat ≤ +2.262

t-stat = (0.7448)*√((11-2)/(1-0.7448²)) 3.3484

Conclusion
Since t-stat (3.3484) lies outside the area of acceptance for H0,there is
sufficient sample evidence at the 5% significance level to reject H0 in favour of H1.
Conclude, at a 5% significance level, that there is a statistically significant association
between opinion poll predictions and the actual election results.

(c) Regression coefficients

b1 = (11*28059 - 541*553)/(11*27789-541²) = 0.729

b0 = (553 - 0.729*541)/11 = 14.4175

Thus ŷ = 14.4175 + 0.729 x where 34 ≤ x ≤ 65

Interpretation of b1 coefficient
For a one percentage point increase in an opinion poll prediction, the actual
election percentage is likely to increase by 0.729 percentage points.

(d) Coefficient of Determination r²

r² = 0.7448² = 0.5547

Interpretation
Opinion poll predictions can explain 55.47% of variation in actual election results percentages
(e) Prediction
x = 58 ŷ= 14.4175 + 0.729*(58) = 56.70%
If an opinion poll predicts support at 58%, the actual election result is likely to be 56.7%.

(f) x = 82 ŷ= 14.4175 + 0.729*(82) = 74.20%


If an opinion poll predicts support at 82%, the actual election result is likely to be 74.2%.
Because x = 82 is beyond the domain of x (34 ≤ x ≤ 65), this expected actual election result
is unreliable and possibly invalid due to extrapolation being used.
Exercise 12.14 File: X12.14 - capital investment.xlsx

(a) Dependent variable (y) = return on investment (%)


Independent variable (x) = capital investment

(b) Scatter Plot

Scatter Plot
Return on Investment (%) vs Capital Investment (%)
10
return on investment (%)

0
10 20 30 40 50 60 70 80 90
levels of capital investm ent (%)

Interpretation
There appears to be a weak to moderate association between a company's level of capital
investment and its return on investment.

(c) Excel's Data Analysis (Regression)

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.41447
R Square 0.17179
Adjusted R Square 0.15253
Standard Error 2.12495
Observations 45

ANOVA
df SS MS F-stat p-value
Regression 1 40.27335 40.27335 8.91907 0.00465
Residual 43 194.16309 4.51542
Total 44 234.43644

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%


Intercept 1.27412 1.19293 1.06806 0.29145 -1.13166 3.67990
Capital 0.06783 0.02271 2.98648 0.00465 0.02203 0.11363

Correlation coefficient (r) = 0,414473

Regression line (equation) ŷ = 1.2741 + 0.06783 x for 21.1 ≤ x ≤ 79.5


(d) Coefficient of Determination r²

r² = 0.41447² = 0.171788

Interpretation
The variation in the level of capital investment explains only 17.18% of the variation
in return on investment .

(e) H0 : ρ = 0 Area of Acceptance for H0


H1 : ρ ≠ 0
t-crit = t(α=0.05,df = 43) = ± 2.016
-2.016 ≤ t-stat ≤ +2.016

t-stat = (0.41447)*√((45-2)/(1-0.41447²)) = 2.9865

Conclusion
Since t-stat (2.9865) lies outside the area of acceptance for H0,there is
sufficient sample evidence at the 5% significance level to reject H0 in favour of H1.
Conclude, at a 5% significance level, that there is a statistically significant association
between a company's level of capital investment and its return on investment.

(f) Interpretation of b1 coefficient b1 = 0.06783


For a one percentage point change in capital investment , company return on investment
can be expected to change by 0.06783 percentage points.

(h) Estimation
x = 55 ŷ = 1.27412 + 0.06783 (55) = 5.005%

The expected return on investment for a company with a 55% level of capital
investment is 5.005%.
Exercise 12.15 File: X12.15 - property valuations.xlsx

(a) Dependent variable (y ) = Market Values


Independent variable (x ) = Council Valuations

Council valuations are assumed to have an influence on property market values.

(b) Scatter Plot

Market Values vs Council Valuations

300

250
market values (R)

200

150

100

50 Market values

0
0 50 100 150 200
council valuations (R)

Interpretation
There appears to be a strong positive correlation between the council's valuation of
a residential property in Bloemfontein and its market value.

(c) Excel's Data Analysis (Regression)

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.78104
R Square 0.61002
Adjusted R Square 0.59975
Standard Error 26.40553
Observations 40

ANOVA
df SS MS F-stat p-value
Regression 1 41444.82 41444.82 59.4402 2.7522E-09
Residual 38 26495.58 697.2521
Total 39 67940.4

Coefficients Standard Error t Stat P-value Lower 95% Upper 95%


Intercept 71.3619 14.2370 5.0124 1.28E-05 42.541 100.183
Council valuations 1.0151 0.1317 7.7097 2.75E-09 0.749 1.282

Correlation coefficient (r) = 0.78104


Regression line (equation) ŷ = 71.3619 + 1.0151 x for 48 ≤ x ≤ 154

(d) Coefficient of Determination r²


r² = 0.78104² = 0.61002
Interpretation
Variation in council valuations explain 61.002% of the variation in property market values .
(e) H0: ρ = 0 Area of Acceptance for H0
H1: ρ ≠ 0
t-crit = t(α=0.05,df = 38) = ± 2.024
-2.024 ≤ t-stat ≤ +2.024

t-stat = (0.78104)*√((40-2)/(1-0.78104²)) = 7.70975

Conclusion
Since t-stat (7.70975) lies well outside the area of acceptance for H0,there is
sufficient sample evidence at the 5% significance level to reject H 0 in favour of H1.
Conclude, at a 5% significance level, that there is a statistically significant association
between the Council's valuation and the resultant property market valuation.

(f) Interpretation of b1 coefficient b1 = 1.0151


For a R1 (in R1000) change (up / down) in council valuation , market valuation of a
property can be expected to change (up / down) by R1.0151 (in R1000) .

(g) Estimation
x = 100 ŷ = 71.3619 + 1.0151 (100) = R172.874 (in R1000s)

The expected market value of a property which the council values at R100 (in R1000s)
is likely to be R172.874 (in R1000s).
CHAPTER 13

MULTIPLE LINEAR REGRESSION

Exercise 13.1 Simple linear regression has only one independent variable (x1) whereas multiple linear
regression has two or more independent variables (x1, x2, x3, … , xk) that are assumed
to influence the outcome of the dependent variable, y .

Exercise 13.2 ANOVA


df SS MS F-stat p-value F-crit
Regression 5 84 16.8 4.541 0.002261 2.45
Residual 40 148 3.7
Total 45 232

(a) R2 = 84/232 = 0.3621


36.2% of total variation in y can be explained by the 5 independent variables.

(b) H0: β1 = β2 = β3 = β4 = β5 = 0 vs H1: At least one βi ≠ 0 (i = 1,2,3,4,5)


or H0: ρ = 0 vs H1: ρ ≠ 0

(c) F-stat = 4.5412 and F-crit = F(0.05,5,40) = 2.45 (See Anova Table)

(d) Reject H0. Conclude that the overall model is statistically significant.
(i.e. at least one x i is statistically significant in estimating y)
Exercise 13.3

Coefficients Std Error t-stat p-value Lower 95% Upper 95%


Intercept 1.82 1.12 1.63 0.1215 -0.53 3.92
A 0.68 0.28 2.44 0.0253 0.09 2.78
B -2.35 0.984 -2.39 0.0140 -4.42 -0.25
C 0.017 0.012 1.42 0.1737 -0.01 2.12
D 1.96 1.16 1.69 0.1083 -0.48 4.06

t-crit = t(0.05,24-4-1) = t(0.05,19) = 2.093

(a) For each x i variable (A, B, C and D), test: H0: βi = 0 against H1: βi ≠ 0 for i = A, B, C and D.

(b), (c), (d) t-crit = t(0.05,19) = ±2.093


For A : Since t-stat (2.44) > t-crit (+2.093); or p -value (0.0253) < α (0.05), or
{0.09 ≤ βA ≤ 2.78} does not cover zero, conclude variable A is statistically significant.

For B : Since t-stat (-2.39) < -t-crit (-2.093); or p -value (0.014) < α (0.05), or
{-4.42 ≤ βB ≤ -0.25} does not cover zero, conclude variable B is statistically significant.

For C : Since t-stat (1.42) < -t-crit (+2.093); or p -value (0.1737) > α (0.05), or
{-0.01 ≤ βC ≤ 2.12} covers zero, conclude variable C is not statistically significant.

For D : Since t-stat (1.69) < -t-crit (+2.093); or p -value (0.1083) > α (0.05), or
{-0.48 ≤ βD ≤ 4.06} covers zero, conclude variable D is not statistically significant.
Exercise 13.4

(a) For x 3 variable, test: H0: β3 = 0 against H1: β3 ≠ 0

(b) t-crit = t(0.05,30) = ±2.042. Hence, do not reject H0 if -2.042 ≤ t-stat ≤ +2.042.

(c ) Since t-stat (2.44) > t-crit (+2.042), hence reject H0 in favour of H1 at α = 0.05.

(d) Conclude that the x 3 variable is statistically significant in estimating y .

Exercise 13.5

(a) Holding all other variables constant, a unit increase in x 2 will result in a 1.6 reduction in y^ .

(b) For x 2 variable, test: H0: β2 = 0 against H1: β2 ≠ 0


Yes, since the 95% confidence interval for β2 does not cover zero.

Exercise 13.6 Binary coding scheme (Choose 'Lean' as the base category)'
F1 and F2 are the dummy variable names chosen.

Fuel type F1 F2
Leaded 1 0
Unleaded 0 1
Lean 0 0

Exercise 13.7 Binary coding scheme (Choose 'spring' as the base category)'
S1, S2 and S3 are the dummy variable names chosen.

Season S1 S2 S3
summer 1 0 0
autumn 0 1 0
winter 0 0 1
spring 0 0 0
Exercise 13.8 File: X13.8 - employee absenteeism.xlsx

Excel's Data Analysis - Regression

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.7384
R Square 0.5453
Adjusted R Square 0.5012
Standard Error 3.7204
Observations 35

ANOVA
df SS MS F-stat p-value
Regression 3 514.467 171.489 12.390 0.00001707
Residual 31 429.076 13.841
Total 34 943.543

Coefficients Standard Error t-stat p-value Lower 95% Upper 95%


Intercept 36.411 5.929 6.141 8.22345E-07 24.318 48.504
Tenure 0.220 0.065 3.376 0.001997 0.087 0.352
Satisfaction -0.184 0.109 -1.692 0.100765 -0.406 0.038
Commitment -0.332 0.080 -4.130 0.000254 -0.497 -0.168

(a) R2 = 514.467/943.543 = 54.53%

(b) H0: βT = βS = βC = 0 versus H1: At least one βi ≠ 0


F-crit = F(0.05,3,31) = 2.92 F-stat = 12.39 and p -value = 0.00001707
Reject H0. Conclude the overall model is statistically significant.
(i.e. at least one x i is statistically significant in estimating y)

(c), (d), (e) For each x i variable, test: H0: βi = 0 against H1: βi ≠ 0 with t-crit = t(0.05,31) = ±2.04
Tenure : Since t-stat (3.376) > t-crit (+2.04); or p -value (0.001997) < α (0.05), or
{0.087 ≤ βT ≤ 0.352} does not cover zero, conclude Tenure is statistically significant.
Satisfaction : Since t-stat (-1.692) lies within t-crit (±2.04); or p -value (0.100765) > α (0.05), or
{-0.406 ≤ βS ≤ 0.038} covers zero, conclude Satisfaction is not statistically significant.
Commitment : Since t-stat (-4.13) < t-crit (-2.04); or p -value (0.000254) < α (0.05), or
{-0.497 ≤ βC ≤ -0.168} does not cover zero, conclude Commitment is statistically significant.

(f) (i) No, organisational commitment is the most important explanatory factor because it has a
larger t-stat value (-4.13) and a smaller p- value (0.000254) than job tenure .

(f) (ii) No, job satisfaction is not a statistically signficant explanatory factor of employee
absenteeism (see (c), (d) and (e) above).
Yes, organisational commitment does play a statistically signficant role in explaining
employee absenteeism (see (c), (d) and (e) above).

(g) y(hat) = 36.411 + 0.22 (48) - 0.184 (50) - 0.332 (60) = 17.8008
Using t-crit = t(0.025,31) = ±2.04; standard error = 3.7204; n = 35
giving margin of error = 2.04 (3.7204)/√35 = 1.2826
Lower 95% confidence limit = 17.8008 - 1.2826 16.5182
Upper 95% confidence limit = 17.8008 + 1.2827 19.0835
{16.52 ≤ y(hat)(estimated) ≤ 19.08}
Management interpretation: We can be 95% confident that the true average number
of days absent per employee per annum is likely to lie between 16.5 days and 19.1 days.
Exercise 13.9 File: X13.9 - plastics wastage.xlsx

Excel's Data Analysis - Regression

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.8061
R Square 0.6498
Adjusted R Square 0.6109
Standard Error 0.5160
Observations 31

ANOVA
df SS MS F-stat p-value
Regression 3 13.3384 4.4461 16.6992 2.466E-06
Residual 27 7.1887 0.2662
Total 30 20.5271

Coefficients Standard Error t-stat p-value Lower 95% Upper 95%


Intercept 1.8179 1.2189 1.4914 0.1474 -0.6830 4.3188
Dexterity -0.1112 0.0286 -3.8816 0.0006 -0.1700 -0.0524
Speed 0.0173 0.0047 3.6770 0.0010 0.0077 0.0270
Viscosity 1.9189 1.2581 1.5252 0.1388 -0.6625 4.5004

(a) R2 = 13.3384/20.5271 = 64.98%

(b) H0: βD = βS = βV = 0 versus H1: At least one βi ≠ 0


F-crit = F(0.05,3,27) = 2.99
F-stat = 16.6992 and p-value = 0.000002466
Reject H0. Conclude the overall model is statistically significant.
(i.e. at least one x i is statistically significant in estimating y)

(c), (d), (e) For each x i variable, test: H0: βi = 0 against H1: βi ≠ 0 with t-crit = t(0.05,27) = ±2.052
Dexterity : Since t-stat (-3.886) < t-crit (-2.052); or p -value (0.0006) < α (0.05), or
{-0.17 ≤ βD ≤ -0.0524} does not cover zero, conclude Dexterity is statistically
significant.
Speed : Since t-stat (3.677) > t-crit (+2.052); or p -value (0.001) < α (0.05), or
{0.0077 ≤ βS ≤ 0.027} does not cover zero, conclude Speed is statistically significant.
Viscosity : Since t-stat (1.525) < t-crit (+2.052); or p -value (0.1388) > α (0.05), or
{-0.6625 ≤ βV ≤ 4.5} covers zero, conclude Viscosity is not statistically significant.

(f) The most important factor is operator dexterity (p -value (0.0006)), then machine speed with
p -value = 0.001. Plastic viscosity is not a significant influencing factor (p -value = 0.1388).

(g) y(hat) = 1.8179 - 0.1112 (25) + 0.0173 (200) + 1.9189 (0.25) = 2.9826
Using t-crit = t(0.025,27) = ±2.052; standard error = 0.516; n = 31
giving margin of error = 2.052 (0.516)/√31 = 0.1902
Lower 95% confidence limit = 2.9826 - 0.1902 = 2.7924
Upper 95% confidence limit = 2.9826 + 0.1902 = 3.1728
{2.79% ≤ y(hat)(estimated) ≤ 3.17%}
Management interpretation: We can be 95% confident that the true average % of plastic
wastage per shift is likely to lie between 2.79% and 3.17%.
Exercise 13.10 File: X13.10 - employee performance.xlsx

(a) Binary coding scheme (Choose 'method C' as the base category)
MA and MB are the dummy variable names chosen.

Method Code MA MB
A 1 0
B 0 1
C 0 0

(b) Sample input data for first 6 consultants (showing the binary coded data)

Consultant Productivity Experience MA MB


1 24 9 1 0
2 30 4 0 1
3 26 10 1 0
4 37 12 0 1
5 29 10 0 0
6 28 6 0 0

Excel's Data Analysis - Regression (using the binary coded data as in (a))

SUMMARY OUTPUT

Regression Statistics
Multiple R 0.7894
R Square 0.6231
Adjusted R Square 0.5524
Standard Error 2.3425
Observations 20

ANOVA
df SS MS F-stat p-value
Regression 3 145.151 48.384 8.817 0.001108
Residual 16 87.799 5.487
Total 19 232.95

Coefficients Std error t-stat p-value Lower 95% Upper 95%


Intercept 26.387 1.940 13.601 3.29E-10 22.274 30.499
Experience 0.389 0.167 2.335 0.0329 0.036 0.742
MA -3.659 1.376 -2.659 0.0172 -6.576 -0.742
MB 1.472 1.336 1.102 0.2867 -1.360 4.304

Estimated multiple regression equation:


y^ = 26.387 + 0.389 Experience - 3.659 MA + 1.472 MB (based on data recoded as in (a))

(c) R2 = 145.151/232.95 = 62.31%

(d) H0: βE = βMA = βMB = 0 versus H1: At least one βi ≠ 0


F-crit = F(0.05,3,16) = 3.24
F-stat = 8.817 and p -value = 0.001108
Reject H0. Conclude the overall model is statistically significant.
(i.e. at least one x i is statistically significant in estimating y)

(e) For Experience , test: H0: βE = 0 against H1: βE ≠ 0 with t-crit = t(0.05,16) = ±2.12
Since t-stat (2.335) > t-crit (2.12), conclude work experience is statistically significant at α = 0.05.

(f) For each of MA and MB, test: H0: βi = 0 vs H1: βi ≠ 0 (i = MA, MB), wiith t-crit = t(0.05,16) = ±2.12

MA : Since t-stat (-2.659) < lower t-crit (-2.12), conclude MA is statistically significant (i.e. adopting
marketing method A results in significantly lower consultant productivity levels, on average,
compared to using marketing method C (i.e. the base category).

MB : Since –t-crit (-2.12) < t-stat (1.102) < +t-crit (-2.12), conclude MB is not statistically significant
(i.e. consultant productivity levels, on average, are the same for marketing method B and
and marketing method C (i.e. no difference to the base category average productivity level).

Overall conclusion: the independent variable ‘marketing method’ is statistically significant, but
only for marketing method A (when compared to method C (i.e. the base category).
Marketing methods B and C can be combined as there is no statistically significant difference
between them with regards to their average productivity levels across consultant.

(g), (h) For each x i variable, test: H0: βi = 0 against H1: βi ≠ 0


Experience : Since its p -value (0.0329) < α = 0.05, or {0.036 ≤ βE ≤ 0.742} does not cover zero,
conclude that work experience is statistically significant.
MA : Since its p -value (0.0172) < α = 0.05, or {-6.576 ≤ βMA ≤ -0.742} does not cover zero,
conclude that marketing method A is statistically significantly different from marketing method C
(i.e. the base category) in terms of consultants' average productivity levels.
MB : Since its p -value (0.2867) > α = 0.05, or {-1.36 ≤ βMA ≤ 4.304} covers zero,
conclude marketing method B is not statistically significantly different from marketing method C
(i.e. the base category) in terms of consultants' average productivity levels.

(i) Employ consultants with longer work experience and avoid using marketing method A as it
produces lower productivity levels than either marketing methods B or C.

(j) y^ = 26.387 + 0.389 (8) - 3.659 (0) + 1.472 (1) = 30.9717


Using t-crit = t(0.025,16) = ±2.1199; standard error = 2.3425; n = 20
giving margin of error = 2.1199 (2.3425)/√20 = 1.1104
Lower 95% confidence limit = 30.9717 - 1.1104 = 29.86
Upper 95% confidence limit = 30.9717 + 1.1104 = 32.08
{29.86 deals ≤ y(hat)(estimated) ≤ 32.08 deals}
Management interpretation: The bank management can be 95% confident that the actual
average number of deals closed per month per consultant is likely to lie between 30 and 32 (rounded).
Exercise 13.11 File: X13.11 - corporate performance.xlsx

(a) Binary coding scheme (Choose region 'KZN' as the base category)
R1 and R2 are the dummy variable names chosen.

Region Code R1 R2
Gauteng 1 1 0
Cape 2 0 1
KZN 3 0 0

Binary coding scheme (Choose sector 'Construction' as the base category)


S is the dummy variable name chosen.

Sector Code S
Agriculture 1 1
Construction 2 0

(b) Sample input data for first 6 companies (showing the binary coded data)

ROC(%) Sales Margin% Debt ratio(%) R1 R2 S


19.7 7178 18.7 28.5 1 0 0
17.2 1437 18.5 24.3 1 0 1
17.1 3948 16.5 65.6 1 0 1
16.6 1672 16.2 26.4 1 0 1
16.6 2317 16.0 20.1 1 0 1
16.5 4123 15.6 46.4 0 1 0

Excel's Data Analysis - Regression (using the binary coded data as in (a))

SUMMARY OUTPUT Y = f(Sales; Margin %; Debt ratio(%); Region; Sector)


Regression Statistics
Multiple R 0.9125
R Square 0.8327
Adjusted R Square 0.7769
Standard Error 0.9524
Observations 25

ANOVA
df SS MS F-stat p-value
Regression 6 81.2595 13.5433 14.9318 0.000004076
Residual 18 16.3261 0.9070
Total 24 97.5856
REGRESSION OUTPUT
Coefficients Standard Error t Stat P-value Lower 95% Upper 95%
Intercept 11.0146 0.8746 12.5936 0.0000 9.1771 12.8520
Sales 0.0002 0.0001 2.1668 0.0439 0.00001 0.00033
Margin% 0.1791 0.0672 2.6656 0.0158 0.03794 0.32030
Debt ratio(%) 0.0091 0.0154 0.5930 0.5606 -0.02321 0.04146
R1 3.1453 0.8389 3.7494 0.0015 1.38288 4.90770
R2 0.9213 0.5862 1.5716 0.1335 -0.31031 2.15295
S -0.9230 0.4205 -2.1950 0.0415 -1.80651 -0.03956

y^ = 11.0146 + 0.0002 Sales + 0.1791 Margin % + 0.0091 Debt ratio % + 3.1453 R1 + 0.9213 R2 - 0.923 S
(based on data recoded as in (a))

2
(c ) R = 81.2595/97.5856 = 83.27%

(d) H0: βS = βM% = βDR = βR1 = βR2 = βS = 0 versus H1: At least one βi ≠ 0
F-crit = F(0.05,6,18) = 2.66
F-stat = 14.932 and p -value = 0.000004076
Reject H0. Conclude that the overall model is statistically significant.
(i.e. at least one x i is statistically significant in estimating y)

(e), (h), (i) For all variables, test: H0: βi = 0 vs H1: βi ≠ 0 with t-crit = t(0.05,18) = ±2.101
Sales : Since t-stat (2.1668) > t-crit (2.101), or p -value (0.0439) < α = 0.05 or
{0.00001 ≤ βS ≤ 0.00033} does not cover zero, conclude sales is statistically significant.
Margin% : Since t-stat (2.6656) > t-crit (2.101), or p -value (0.0158) < α = 0.05 or
{0.0379 ≤ βM% ≤ 0.3203} does not cover zero, conclude margin% is statistically significant.
Debt ratio% : Since –t-crit (-2.101) < t-stat (0.593) < t-crit (2.101), or p -value (0.5606) > α = 0.05 or
{-0.0232 ≤ βDR% ≤ 0.0415} covers zero, conclude debt ratio% is not statistically significant.

(f) Region: For each dummy variable R1 and R2 , test: H0: βi = 0 against H1: βi ≠ 0 with t-crit = t(0.05,18) = ±2.101
R1 (Gauteng) : Since t-stat (3.7494) > upper t-crit (2.101); or its p -value (0.0015) < α = 0.05, or
{1.38288 ≤ βR1 ≤ 4.9077} does not cover zero, conclude that companies that operate in the Gauteng region
have a statistically significantly higher return on capital (%), on average, than companies that operate
in the KZN region (i.e. the base region).
R2 (Cape) : Since t-stat (1.5716) < upper t-crit (2.101); or its p -value (0.1335) > α = 0.05, or
{-0.31031 ≤ βR2 ≤ 2.15295} covers zero, conclude that companies that operate in the Cape region
do not have a statistically significant difference in their average return on capital (%)than companies
that operate in the KZN region (i.e. the base region).
Overall conclusion: the independent variable ‘region’ is statistically significant, but only with
respect to the Gauteng region (when compared to th KZN region (i.e. the base region).
The Cape and KZN regions can be merged into a single region as there is no statistically significant
difference in the average return on capital (%) of companies operating within these two regions.

(g) Sector: For the dummy variable, S, test: H0: βS = 0 against H1: βS ≠ 0 with t-crit = t(0.05,18) = ±2.101
S (Agriculture) : Since t-stat (-2.195) < lower t-crit (-2.101); or its p -value (0.0415) < α = 0.05, or
{-1.80651 ≤ βS ≤ -0.03956} does not cover zero, conclude that companies that operate in the
agricultural sector have a statistically significantly lower return on capital (%), on average, than
companies that operate in the construction sector (i.e. the base sector).
Overall conclusion: the independent variable ‘sector’ is statistically significant.
(j) Significant performance measures of ROC% are: Sales, Margin%, but not Debt ratio%.
For region, Gauteng has a significantly positive impact on average ROC% compared to Cape and KZN.
The agricultural sector has a significantly negative impact on average ROC% compared to the construction sector.

(j) y^ = 11.0146 + 0.0002 (8862) + 0.1791 (10) + 0.0091 (22) + 3.1453 (0) + 0.9213 (1) - 0.923 (0) = 15.6995
Using t-crit = t(0.025,18) = ±2.101; standard error = 0.9524; n = 25
giving margin of error = 2.101 (0.9524)/√25 = 0.400198
Lower 95% confidence limit = 15.6995 - 0.400198 = 15.299
Upper 95% confidence limit = 15.6995 + 0.400198 = 16.100
{15.299% ≤ y(hat)(estimated) ≤ 16.10%}
Management interpretation: The investment analyst can be 95% confident that the actual
average return on capital (%) of companies with the given profile lies between 15.3% and 16.1%.
CHAPTER 14

INDEX NUMBERS

MEASURING BUSINESS ACTIVITY

Exercise 14.1 An index number is a single summary value that measures the overall
change in the level of activity of a single item or a basket of related
items from one time period to the another.

Example: Consumer Price Index (CPI) - the inflation indicator.

Example 14.2 A price index measures changes in price levels over time, holding
quantities constant.
A quantity index measures changes in consumption levels over time, holding
prices constant.

Example 14.3 Items are 'weighted' in a basket to reflect the importance (or value) of each
item in the basket relative to the other items in the basket.

Example 14.4 Laspeyres weighting method and Paasche weighting method

Example 14.5 (i) The purpose (or scope) of the index


(ii) The selection of the basket of items (i.e. the mix of items)
(iii) The choice of item weights
(iv) The choice of a suitable base year
(v) The formulation of a substitution rule

Example 14.6 A link relative is a period-on-period (consecutive) change in level of activity.


A price relative is a change in the level of activity of an item in a given period
relative to a base period.

Example 14.7 Real (constant) values are found by dividing monetary values by an 'inflation' index.
This removes the influence of price increases.
Real (constant) values refer to the actual purchasing power of money / or the
real (actual) change in the level of activity.
Exercise 14.8 File: X14.8 - motorcycle sales.xlsx

Data 2009 2010


Motorcycle Unit price Quantity Unit price Quantity
model (R1000) (units sold) (R1000) (units sold)
A 25 10 30 7
B 15 55 19 58
C 12 32 14 40

Motorcycle Price
(a) p1/p0*100
model Relative
A 30/25*100 = 120.0
B 19/15*100 = 126.7
C 14/12*100 = 116.7

Interpretation
Model A has risen in price by 20% from 2009 to 2010; model B by 26.7%
and model C by 16.7%.

(b) (i) Laspeyres Weighted Aggregates Price Index

Motorcycle
Base Value (p0*q0) Current Value (p1*q0)
model
A 25*10 = 250 30*10 = 300
B 15*55 = 825 19*55 = 1045
C 12*32 = 384 14*32 = 448
Totals 1459 1793

Composite Price Index 1793/1459*100 = 122.9

(b) (ii) Laspeyres Weighted Average of Price Relatives Index

Weighted
Motorcycle
Base Value (p0*q0) Price Relative (p1/p0) Price
model
Relatives
A 25*10 = 250 30/25*100 = 120.0 120*250 +
B 15*55 = 825 19/15*100 = 126.7 126.7*825 +
C 12*32 = 384 14/12*100 = 116.7 116.7*384 =
Totals 1459 179300

Composite Price Index 179300/1459 = 122.9

(c) Interpretation
Motorcycle models A, B and C have risen in price by 22.9% on average
from 2009 to 2010.
Exercise 14.9 File: X14.8 - motorcycle sales.xlsx

Data 2009 2010


Motorcycle Unit price Quantity Unit price Quantity
model (R1000) (units sold) (R1000) (units sold)
A 25 10 30 7
B 15 55 19 58
C 12 32 14 40

Motorcycle Quantity
(a) q1/q0*100
model Relative
A 7/10*100 = 70.0
B 58/55*100 = 105.5
C 40/32*100 = 125.0

Interpretation
Model A unit sales dropped 30% from 2009 to 2010; model B unit sales rose
by 5.55%; while model C unit sales rose the most by 25% over the year.

(b) (i) Laspeyres Weighted Aggregates Quantity Index

Motorcycle
Base Value (p0*q0) Current Value (p0*q1)
model
A 25*10 = 250 25*7 = 175
B 15*55 = 825 15*58 = 870
C 12*32 = 384 12*40 = 480
Totals 1459 1525

Composite Quantity Index 1525/1459*100 = 104.5

(b) (ii) Laspeyres Weighted Average of Quantity Relatives Index

Weighted
Motorcycle
Base Value (p0*q0) Quantity Relative (q1/q0) Quantity
model
Relatives
A 25*10 = 250 7/10*100 = 70.0 70*250 +
B 15*55 = 825 58/55*100 = 105.5 105.5*825 +
C 12*32 = 384 40/32*100 = 125.0 125*384 =
Totals 1459 152537.5

Composite Quantity Index 152537.5/1459 = 104.5

(c) Interpretation
Unit sales of motorcycle models A, B and C have risen by 4.5% on average
from 2009 to 2010.
Exercise 14.10 File: X14.8 - motorcycle sales.xlsx

Data 2009 2010


Motorcycle Unit price Quantity Unit price Quantity
model (R1000) (units sold) (R1000) (units sold)
A 25 10 30 7
B 15 55 19 58
C 12 32 14 40

(a) (i) Paasche Weighted Aggregates Quantity Index

Motorcycle
Base Value (p1*q0) Current Value (p1*q1)
model
A 30*10 = 300 30* 7 = 210
B 19*55 = 1045 19*58 = 1102
C 14*32 = 448 14*40 = 560
Totals 1793 1872

Composite Quantity Index 1872/1793*100 = 104.41

(a) (ii) Paasche Weighted Average of Quantity Relatives Index

Weighted
Motorcycle
Base Value (p1*q0) Quantity Relative (q1/q0) Quantity
model
Relatives
A 30*10 = 300 7/10*100 = 70.0 70*300 +
B 19*55 = 1045 58/55*100 = 105.5 105.5*1045 +
C 14*32 = 448 40/32*100 = 125.0 125*448 =
Totals 1793 187200

Composite Quantity Index 187200/1793 = 104.41

(b) Interpretation
Unit sales of motorcycle models A, B and C have risen by 4.41% on average
from 2009 to 2010.
Exercise14.11 File: X14.11 - Telkom services.xlsx

Data 2009 2010 2011


Telkom Services Unit price Quantity Unit price Quantity Unit price Quantity
(cents/call) (100's calls) (cents/call) (100's calls) (cents/call) (100's calls)
TalkPlus 65 14 70 18 55 17
SmartAccess 35 27 40 29 45 24
ISDN 50 16 45 22 40 32

(a) Telkom Services Price relatives (2010) Price relatives (2011)


TalkPlus 70/65*100 = 107.7 55/65*100 = 84.6
SmartAccess 40/35*100 = 114.3 45/35*100 = 128.6
ISDN 45/50*100 = 90.0 40/50*100 = 80.0

Interpretation
TalkPlus services increased by 7.7% in price from 2009 to 2010, but then dropped by
15.4% in price from 2009 to 2011.
SmartAccess on the other hand showed an increase in price from 2009 to 2010 of 14.3%
and by 28.6% from 2009 to 2011.
ISDN showed a decrease in price by 10% in the first year (from 2009 to 2010) and by 20%
over the 2 year period from 2009 to 2011.

(b) Laspeyres Weighted Aggregates Price Index

Base Value 2010 Value 2011 Value


Telkom Services
(p0*q0) (p1*q0) (p1*q0)
TalkPlus 910 980 770
SmartAccess 945 1080 1215
ISDN 800 720 640
2655 2780 2625

Composite Price Indexes 104.71 98.87

Interpretation (based on the Laspeyres approach)


The cost of Telkom services increased, on average, by 4.71% from 2009 to 2010,
while there was a net reduction in costs, on average of 1.13% from 2009 to 2011.

(c) Paasche Weighted Aggregates Quantity Index

2010 Prices 2011 Prices


Base Value 2010 Value Base Value 2011 Value
Telkom Services
(p1*q0) (p1*q1) (p1*q0) (p1*q1)
TalkPlus 980 1260 770 935
SmartAccess 1080 1160 1215 1080
ISDN 720 990 640 1280
2780 3410 2625 3295

Composite Quantity Indexes 122.7 125.5

(d) Interpretation (based on the Paasche approach)


The printing company's usage of Telkom services has risen by 22.7% over one
year (from 2009 to 2010) and by 25.5%, on average over two years (from 2009 to 2011).
Exercise 14.12 File: X14.12 - computer personnel.xlsx

Data Annual salary (in R10 000) No. of employees


Job categories
2008 2011 2008 2011
Systems analyst 42 50 84 107
Programmer 29 36 96 82
Network manager 24 28 58 64

(a) Laspeyres Weighted Aggregate Salary Index

Base Value Current Value


IT Job Category
2008 (p0*q0) 2011 (p1*q0)
Systems analyst 3528 4200
Programmer 2784 3456
Network manager 1392 1624
7704 9280

Laspeyres Composite Salary Index 120.46

(b) Interpretation On average, the overall remuneration has increased by


20.46% over the 3 years period from 2008 to 2011.

Price relatives
(c) IT Job Category
(p1/p0)
Systems analyst 119.05
Programmer 124.14
Network manager 116.67

Interpretation Programmers have enjoyed the largest percentage increase


in remuneration of 24.14% from 2008 to 2011.
Exercise 14.13 File: X14.12 - computer personnel.xlsx

Data Annual salary (in R10 000) No. of employees


Job categories
2008 2011 2008 2011
Systems analyst 42 50 84 107
Programmer 29 36 96 82
Network manager 24 28 58 64

Staff Relatives
(a) IT Job Category
(q1/q0)
Systems analyst 127.4
Programmer 85.4
Network manager 110.3

Interpretation
The staff complement of Systems Analysts grew by 27.4% and that of Network
managers grew by 10.3% over the period from 2008 to 2011. However, the
number of Programmers, on the other hand, reduced by 14.6% over this same period.

(b) (i) Laspeyres Quantity index (Weighted Aggregates approach)

Base Value Current Value


IT Job Category
2008 (p0*q0) 2011 (p0*q1)
Systems analyst 3528 4494
Programmer 2784 2378
Network manager 1392 1536
7704 8408

Composite Staff Complement (Quantity) Index 109.14

(b) (ii) Laspeyres Quantity index (Weighted Average of Relatives approach)

Weighted
Base Value 2008 Staff Relatives Ave of
IT Job Category
(p0*q0) (q1/q0) Quantity
Relatives
Systems analyst 3528 127.38 449400
Programmer 2784 85.42 237800
Network manager 1392 110.34 153600
7704 840800

Composite Staff Complement (Quantity) Index 109.14

(c) Interpretation
The overall IT staff complement across all job categories has increased by an
average of 9.14% from 2008 to 2011.
Exercise 14.14 File: X14.14 - printer cartridges.xlsx

Data

2008 2009 2010


Printer
cartridges
Quantity Quantity
Unit price Unit price Quantity used Unit price
used used
HQ21 145 24 155 28 149 36
HQ25 172 37 165 39 160 44
HQ26 236 12 255 12 262 14
HQ32 314 10 306 8 299 11

Print Unit price Unit price


(a) Price relative 2010
Cartridges (2008) (2010)
HQ26 236 262 262/236% = 111.02
HQ32 314 299 299/314% = 95.22

Interpretation
The price of the HQ26 printer cartridge has increased by 11.02% from 2008 to 2010.
The price of the HQ32 printer cartridge has decreased by 4.78% from 2008 to 2010.

(b) (i) Paasche Weighted Aggregates Price Index

2009 Prices 2010 Prices


Printer Base Value Current Value Base Value Current Value
cartridges (2008) (2009) (2008) (2010)
HQ21 4060 4340 5220 5364
HQ25 6708 6435 7568 7040
HQ26 2832 3060 3304 3668
HQ32 2512 2448 3454 3289
Totals 16112 16283 19546 19361

Composite Price Indexes 101.06 99.05

(b) (ii) Paasche Weighted Average of Relatives Price Index - for 2009

Printer Base Value Price relative Weighted Ave


cartridges (2008) (2009) (2009)
HQ21 4060 106.9 434000
HQ25 6708 95.9 643500
HQ26 2832 108.1 306000
HQ32 2512 97.5 244800
Totals 16112 1628300

Paasche Composite Price Index = 101.06

Paasche Weighted Average of Relatives Price Index - for 2010


Printer Base Value Price relative Weighted Ave
cartridges (2008) (2010) (2010)
HQ21 5220 102.8 536400
HQ25 7568 93.0 704000
HQ26 3304 111.0 366800
HQ32 3454 95.2 328900
Totals 19546 1936100

Paasche Composite Price Index 99.05

(c) Interpretation
The average price of print cartridges increased marginally by 1.06% from 2008 to 2009.
However, the average price of print cartridges decreased by 0.95% from 2008 to 2010.

(d) Composite Link Relatives


2008 2009 2010
Composite price indexes 100 101.06 99.05
Composite link relatives 101.06 98.01
Exercise 14.15 File: X14.15 - electrical goods.xlsx

Data 2004 2005 2006 2007 2008 2009 2010


Composite Price Index
88 96 100 109 114 112 115
(Electrical goods)

(a) Reset base year to 2009 100k/112

2004 2005 2006 2007 2008 2009 2010


Composite Price Index
78.6 85.7 89.3 97.3 101.8 100 102.7
(Electrical goods)

Interpretation
In 2004, the average price of electrical goods was 21.4% below the 2009 (base) price levels,
while in 2006, it was only 10.7% below the 2009 (base) price levels.
In 2010, prices were 2.7% higher on average than in the base period of 2009.

(b) Link relatives

2004 2005 2006 2007 2008 2009 2010


Composite Price Index
100 109.1 104.2 109 104.6 98.2 102.7
(Electrical goods)

Interpretation
The annual average price changes in electrical goods starting in 2004 was 9.1% (for 2005);
4.2% (for 2006); 9% (for 2007); 4.6% (for 2008); -1.8% (decrease in 2009); and 2.7% (for 2010).
Exercise 14.16 File: X14.16 - insurance claims.xlsx

Data 2006 2007 2008 2009 2010 2011


Federal Insurance
92.3 95.4 100 102.6 109.4 111.2
(base = 2008)
Baltic Insurance
93.7 101.1 98.2 100 104.5 107.6
(base = 2009)

(a) 2006 2007 2008 2009 2010 2011


Federal Insurance
84.4 87.2 91.4 93.8 100 101.6
(base = 2010)
Baltic Insurance
89.7 96.7 94 95.7 100 103
(base = 2010)

(b) Ferderal Insurance showed a 8.6/91.4% (9.41%) increase from 2008 to 2010.
Baltic Insurance showed a 6.0/94% (6.4%) increase from 2008 to 2010.
Hence Federal Insurance showed the bigger claims increase from 2008 to 2010.

(c) Link Relatives

2006 2007 2008 2009 2010 2011


Federal Insurance 100 103.4 104.8 102.6 106.6 101.6
Baltic Insurance 100 107.9 97.1 101.8 104.5 103

(d) Baltic showed a year on year increase of 3%, while


Federal showed a year on year increase of only 1.6%. Hence Baltic Insurance.

5
(e) Geometric mean Federal √(1.034*1.048*1.026*1.066*1.016) -1 = 3.785%
5
Baltic √(1.079*0.971*1.018*1.045*1.03) - 1 = 2.799%

(f) Federal Insurance's claims processed increased by an average of 3.785% annually


between 2006 and 2011.
Exercise 14.17 File: X14.17 - micro-market basket.xlsx

Data Food items in micro Unit prices (in Rand) Consumption


market basket 2010 2011 2010 2011
Milk (litres) 7.29 7.89 117 98
Bread (loaves) 4.25 4.45 56 64
Sugar (kg) 2.19 2.45 28 20
Maize meal (kg) 5.25 5.59 58 64

(a) Food items 2010 2011 Price Relatives


Milk (litres) 7.29 7.89 108.23
Bread (loaves) 4.25 4.45 104.71
Sugar (kg) 2.19 2.45 111.87
Maize meal (kg) 5.25 5.59 106.48

Interpretation
The price of milk rose by 8.23%; bread by 4.71%; sugar by 11.87% and
maize meal by 6.48% per unit of measure from 2010 to 2011.
The largerst price change (increase) was sugar with a 11.87% increase.

(b) Paasche Weighted Average of Price Relatives Index

Base Value Weighted


Food items Price Relatives
(p0*q1) Average
Milk (litres) 714.42 108.23 77322
Bread (loaves) 272 104.71 28480
Sugar (kg) 43.8 111.87 4900
Maize meal (kg) 336 106.48 35776
Total 1366.22 146478

Paasche Composite Price Index 146478/1366.22 = 107.21

On average, the price of the micro-basket of items increased by 7.21% from 2010 to 2011.

Quantity
(c) Food items 2010 2011
Relatives
Milk (litres) 117 98 83.8
Bread (loaves) 56 64 114.3
Sugar (kg) 28 20 71.4
Maize meal (kg) 58 64 110.3

Interpretation
The consumption of milk decreased by 16.2%; bread consumption rose by 14.3%;
sugar consumption decreased significantly by 28.6% and maize meal consumption
rose by 10.3% from 2010 to 2011 respectively.
The largest consumption change (decrease) was sugar with a 28.6% decrease.
It is interesting to note that sugar showed the largest unit price increase while
simultaneously recorded the largest decrease in consumption from 2010 to 2011.
(d) Paasche Weighted Average of Quantity (Consumption) Relatives Index

Base Value Weighted


Food items Quantity Relatives
(p1*q0) Average
Milk (litres) 923.13 83.8 77322
Bread (loaves) 249.20 114.3 28480
Sugar (kg) 68.60 71.4 4900
Maize meal (kg) 324.22 110.3 35776
Total 1565.15 146478

Paasche Composite Quantity Index 146478/1565.15 = 93.6

On average, consumption of the micro-basket items dropped by 6.4% from 2010 to 2011.
Exercise 14.18 File: X14.18 - utilities usage.xlsx

Data Prices (in Rand / unit) Consumption (No. of units)


Utilities
2008 2009 2010 2008 2009 2010
Electricity 1.97 2.05 2.09 745 812 977
Sewage 0.62 0.68 0.72 68 56 64
Water 0.29 0.31 0.35 296 318 378
Telephone 1.24 1.18 1.06 1028 1226 1284

Price
Unit Price Unit Price
(a) Utilities relative
2008 2010
2010
Electricity 1.97 2.09 106.1
Sewage 0.62 0.72 116.1
Water 0.29 0.35 120.7
Telephone 1.24 1.06 85.5

Interpretation
From 2008 to 2010, electricity increased by 6.1%; sewage costs by 16.1%
water by 20.7%; while telephone costs decreased by 14.5% over this period.
Electricity showed the smallest change (increase) of 6.1% from 2008 to 2010.

(b) (i) Laspeyres Weighted Aggregates Price Index

Current Current
Base Value
Utilities Value Value
(2008)
(2009) (2010)
Electricity 1467.7 1527.3 1557.1
Sewage 42.2 46.2 49.0
Water 85.8 91.8 103.6
Telephone 1274.7 1213.0 1089.7
Totals 2870.4 2878.3 2799.3

Laspeyres Composite Price Indexes 100.3 97.5

(b) (ii) Laspeyres Weighted Average of Price Relatives Index

Price Price
Base Value Weighted Weighted
Utilities relative relative
(2008) Ave (2009) Ave (2010)
(2009) (2010)
Electricity 1467.7 104.1 106.1 152782.4 155717.7
Sewage 42.2 109.7 116.1 4625.0 4894.8
Water 85.8 106.9 120.7 9176.3 10360.9
Telephone 1274.7 95.2 85.5 121353.3 108988.6
Totals 2870.4 287937.0 279961.9

Laspeyres Composite Price Indexes 100.3 97.5


(c) Interpretation
There was virtually no change in the average cost of household utilities between
2008 and 2009. A slight increase of only 0.3% was recorded.
From 2008 to 2010, however, the average cost of household utilities actually
decreased marginally by 2.5%.
Exercise 14.19 File: X14.18 - utilities usage.xlsx

Data Prices (in Rand / unit) Consumption (No. of units)


Utilities
2008 2009 2010 2008 2009 2010
Electricity 1.97 2.05 2.09 745 812 977
Sewage 0.62 0.68 0.72 68 56 64
Water 0.29 0.31 0.35 296 318 378
Telephone 1.24 1.18 1.06 1028 1226 1284

Quantity
Usage Usage
(a) Utilities relative
2008 2010
2010
Electricity 745 977 131.1
Sewage 68 64 94.1
Water 296 378 127.7
Telephone 1028 1284 124.9

Interpretation
Consumption of electricity, water and telephone increased by 31.1%; 27.7% and
24.9% respectively from 2008 to 2010. Only sewage showed a decline in usage
by 5.9% from 2008 to 2010.

(b) (i) Laspeyres Weighted Aggregates Quantity (Consumption) Index

Current Current
Base Value
Utilities Value Value
(2008)
(2009) (2010)
Electricity 1467.7 1599.6 1924.7
Sewage 42.2 34.7 39.7
Water 85.8 92.2 109.6
Telephone 1274.7 1520.2 1592.2
Totals 2870.4 3246.8 3666.2

Laspeyres Composite Quantity Indexes 113.1 127.7

(b) (ii) Laspeyres Weighted Average of Quantity Relatives Index

Quantity Quantity
Base Value Weighted Weighted
Utilities relative relative
(2008) Ave (2009) Ave (2010)
(2009) (2010)
Electricity 1467.7 109.0 131.1 159959 192468
Sewage 42.2 82.4 94.1 3474 3968
Water 85.8 107.4 127.7 9219 10962
Telephone 1274.7 119.3 124.9 152074 159213
Totals 2870.4 324726 366610

Laspeyres Composite Quantity Indexes 113.1 127.7

(c) Interpretation
On average, household consumption of utilities increased by 13.1% in 2009 from 2008
and also showed an overall consumption increase of 27.7% from 2008 to 2010.
Exercise 14.20 File: X14.20 - leather goods.xlsx

Data 2005 2006 2007 2008 2009 2010 2011


Composite
97 92 100 102 107 116 112
cost index

(a) Re-base to 2009 Adjustment factor = 100/107 = 0.934579

2005 2006 2007 2008 2009 2010 2011


Composite
90.7 86 93.5 95.3 100 108.4 104.7
cost index

(b)
Composite cost index

120

110

100
Cost Index

90

80

70 Composite cost index

60
2005 2006 2007 2008 2009 2010 2011

Interpretation
Prior to 2009, the average annual cost of leather goods inputs were below the 2009 level,
but showed a steady increase towards 2009 prices over this period.
Relative to 2009 unit costs, the cost of leather
goods inputs was higher by 8.4% and 4.7% respectively for 2010 and 2011.

(c) The average increase in unit costs of leather goods inputs was only 4.7% between
2009 and 2011.

(d) Link Relatives

2005 2006 2007 2008 2009 2010 2011


Composite
100 94.8 108.7 102.0 104.9 108.4 96.6
cost index

The largest year-on-year change in overall unit costs of leather goods inputs was
between 2006 and 2007 with an increase of 8.7%.
6
(e) Geometric mean √(0.948*1.087*1.02*1.049*1.0841*0.966)
1.024244
(i.e. 2.424% increase on average per year)

=GEOMEAN(0.9485,1.087,1.02,1.049,1.0841,0.9655)
= 1.024244
Exercise 14.21 File: X14.21 - accountants' salaries.xlsx

(a) Year Salary CPI Real Salary


2005 387 95 407.4
2006 406 100 406.0
2007 422 104 405.8
2008 448 111 403.6
2009 466 121 385.1
2010 496 126 393.7
2011 510 133 383.5

(b) Interpretation
The base real salary is R406 (in R1000s). Since 2006, real salaries have
declined relative to 2006 (base) and are continuing to fall further behind inflation (CPI).

(c) Link Relatives


Year Salary CPI Salary CPI Diff (Sal - CPI)
2005 387 95
2006 406 100 104.9 105.3 -0.4
2007 422 104 103.9 104.0 -0.1
2008 448 111 106.2 106.7 -0.6
2009 466 121 104.0 109.0 -5.0
2010 496 126 106.4 104.1 2.3
2011 510 133 102.8 105.6 -2.7

Interpretation
On a year-on-year basis, salary increases have lagged behind the inflation rate (CPI)
in all years except 2010 when salary adjustments exceeded CPI by 2.3%.
Exercise 14.22 File: X14.22 - school equipment .xlsx

Data 2003 2004 2005 2006 2007 2008 2009 2010


Composite
94.8 97.6 100 105.2 108.5 113.9 116.7 121.1
cost index

(a) Re-base to 2008 Adjustment factor = 100/113.9 = 0.878

2003 2004 2005 2006 2007 2008 2009 2010


Composite
83.2 85.7 87.8 92.4 95.3 100 102.5 106.3
cost index

(b) and (e) Plot and Interpretation of Re-based Cost Index Series

Composite Cost Index

110

105

100

95

90

85

80 Composite cost index

75
2003 2004 2005 2006 2007 2008 2009 2010

Interpretation
The average cost of school equipment has shown a steady increase of 3% to 5%
annually over the period 2003 to 2010.

(c) Link Relatives

2003 2004 2005 2006 2007 2008 2009 2010


Composite
100 102.95 102.46 105.20 103.14 104.98 102.46 103.77
price index

Interpretation
The link relatives confirm the year-on-year increases in the average cost of school
equipment of between 3% to 5% p.a.

(d) The 2003 budget must be adjusted annually by the link relative (year-on-year) indexes.
Therefore multiply each previous year's budget by the next year's link relative index.

2003 2004 2005 2006 2007 2008 2009 2010


Budgets 5000000 5147500 5274129 5548384 5722603 6007589 6155376 6387434
Exercise 14.23 File: X14.23 - coffee imports.xlsx

Unit price Quantity Unit price Quantity


Data Coffee types
(2007) (2007) (2010) (2010)
Java 85 52 98 46
Colombia 64 75 74 90
Sumatra 115 18 133 20
Mocha 38 144 42 168

Unit price Unit price


(a) Coffee types Price Relatives
(2007) (2010)
Java 85 98 115.3
Colombia 64 74 115.6
Sumatra 115 133 115.7
Mocha 38 42 110.5

Interpretation
The coffee types of Java, Colombia and Sumatra increased by over 15% from
2007 to 2010, while Mocha increased by only 10.5% over this same period.

(b) Laspeyres Composite Price Index (Weighted Aggregates approach)

Base Value Current Value


Coffee type
(2007) (2010)
Java 4420 5096
Colombia 4800 5550
Sumatra 2070 2394
Mocha 5472 6048
Total 16762 19088

Laspeyres Composite Price Index = 19088/16762 = 113.9

Interpretation
The cost of coffee imports has increased by 13.9% on average from 2007 to 2010.

(c) Cheaper.
The cost of coffee imports has increased by only 13.9% on average, since 2007.
(d) Laspeyres Composite Quantity Index (Weighted Aggregates approach)

Base Value Current Value


Coffee type
(2007) (2010)
Java 4420 3910
Colombia 4800 5760
Sumatra 2070 2300
Mocha 5472 6384
Total 16762 18354

Laspeyres Composite Quantity Index 18354/16762 = 109.5

Interpretation
The quantity of coffee imported has increased by 9.5% on average from 2007 to 2010.
Exercise 14.24 File: X14.24 - medical claims.xlsx

Ave Value (R) Claims Ave Value (R) Claims


Data Claim Type
(2008) (2008) (2010) (2010)
GPs 220 20 255 30
Specialists 720 30 822 25
Dentists 580 10 615 15
Medicines 400 50 438 70

(a) Laspeyres Composite Quantity Index (Weighted Aggregates approach)


(Use the Laspeyres Approach as the default method if none is specified)

Base Value Current Value


Claim type
(2008) (2010)
GP's 4400 6600
Specialists 21600 18000
Dentists 5800 8700
Medicines 20000 28000
Total 51800 61300

Laspeyres Composite Quantity Index = 61300/51800 = 118.34

Interpretation
The number of claims received has increased by 18.34% from 2008 to 2010.

Quantity
(b) Claim type Claims (2008) Claims (2010)
relatives
GP's 20 30 150.0
Specialists 30 25 83.33
Dentists 10 15 150.0
Medicines 50 70 140.0

Interpretation
GP's and Dentists showed the biggest increase in number of claims (both 50% increase).

(c) Laspeyres Composite Price (Claims Value) Index (Weighted Aggregates approach)

Base Value Current Value


Claim type
(2008) (2010)
GP's 4400 5100
Specialists 21600 24660
Dentists 5800 6150
Medicines 20000 21900
Total 51800 57810

Laspeyres Composite Price Index = 57810/51800 = 111.6

(d) Interpretation
The value of claims received between 2008 and 2010 has risen by 11.6%.
Exercise 14.25 File: X14.25 - tennis shoes.xlsx

Unit price Pairs Sold Unit price Pairs Sold


Data Shoe Model
(2009) (2009) (2010) (2010)
Trainer 320 96 342 110
Balance 445 135 415 162
Dura 562 54 595 48

(a) Quantity Relatives

Pairs sold Pairs sold Quantity


Shoe model
(2009) (2010) relatives
Trainer 96 110 114.6
Balance 135 162 120.0
Dura 54 48 88.9

Interpretation
Dura's sales volume is down by 11.1% from 2009.

(b) Laspeyres Weighted Aggregates Quantity (Volume) Index

Base Value Current Value


Shoe model
(2009) (2010)
Trainer 30720 35200
Balance 60075 72090
Dura 30348 26976
Total 121143 134266

Composite Quantity Index = 134266/121143*100 = 110.83

(c) Interpretation
The average increase in shoes sold from 2009 to 2010 was only 10.83%.
Overall, sales volumes do not meet the required growth of at least 12% p.a.
Exercise 14.26 File: X14.26 - energy fund.xlsx

Data
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Fuel
cost 100.0 116.2 122.4 132.1 135.7 140.3 142.8 146.9 153.4 160.5
index

(a) Re-base to 2006 Adjustment factor = 100/140.3 = 0.7128

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Fuel
cost 71.28 82.83 87.25 94.16 96.73 100.00 101.79 104.71 109.34 114.40
index

(b) Link Relatives

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Fuel
cost 100.00 116.20 105.34 107.92 102.73 103.39 101.78 102.87 104.42 104.63
index

Fuel Cost Index

120.00

115.00

110.00

105.00

100.00

95.00
Fuel cost index
90.00
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
(c) Geometric Mean 9√(1.162*1.0534*1.0792*1.0273*1.0339*1.0178*1.0287*1.0442*1.0463)
1.05397
Annual Average Increase = 5.3974% p.a. on average.

Using Excel =GEOMEAN(all link relatives) 105.3976

(d) Interpretation
Annual fuel cost increases and decreases were high in the period 2001 to 2004.
Thereafter, year-on-year increases were moderate and stable at between 3% to 5% p.a.
Exercise 14.27 File: X14.27 - motorcycle distributor.xlsx

Average Selling Price (R)


Data Model 2007 2008 2009 2010 2011
Blitz 18050 19235 21050 21950 22400
Cruiser 25650 26200 27350 28645 31280
Classic 39575 42580 43575 43950 46750

Units Sold
Model 2007 2008 2009 2010 2011
Blitz 205 185 168 215 225
Cruiser 462 386 402 519 538
Classic 88 70 111 146 132

(a) Price Relative Index series - Cruiser model

Model 2007 2008 2009 2010 2011


Cruiser 100.0 102.1 106.6 111.7 121.9

The price of the Cruiser model rose marginally (between 2% and 7%) for 2008
and 2009 relative to 2007; but rose strongly by 11.7% in 2010 and 21.9% in 2011.

(b) Laspeyres Composite Price Index (Using Weighted Aggregates method)

Base Value Value Value


Model Value 2008
2007 2009 2010 2011
Blitz 3700250 3943175 4315250 4499750 4592000
Cruiser 11850300 12104400 12635700 13233990 14451360
Classic 3482600 3747040 3834600 3867600 4114000
Total 19033150 19794615 20785550 21601340 23157360

2007 2008 2009 2010 2011


Price Index 100 104.0 109.2 113.5 121.7

Laspeyres Composite Price Index - Motorcycles


130

120

110

100

90

80

70 Price Index

60
2007 2008 2009 2010 2011
(c) Interpretation
There has been a reasonably constant increase in the average price of motorcycles
over the past 5 years. In 2011, motorcycles cost 21.7% more on average than
they did in 2007.

(d) Link relative of Selling Prices - Classic Model

2007 2008 2009 2010 2011


Classic 100 107.6 102.3 100.9 106.4

Interpretation
The Classic model has shown a very modest year-on-year price increase of
between 2.3% and 7.6% between 2007 and 2011. In 2010, the annual
(year-on-year) increase was only 0.9% (less than 1%).

4
(e) Geometric Mean = √ (1.076*1.023*1.009*1.064) - 1= 4.263
Using Excel = GEOMEAN(link relatives) - 100 = 4.253

Motorcycle prices have risen by an average of 4.263% per annum.

(f) Laspeyres Composite Quantity Index (Using Weighted Aggregates method)

Base Value Value Value


Model Value 2008
2007 2009 2010 2011
Blitz 3700250 3339250 3032400 3880750 4061250
Cruiser 11850300 9900900 10311300 13312350 13799700
Classic 3482600 2770250 4392825 5777950 5223900
Total 19033150 16010400 17736525 22971050 23084850

2007 2008 2009 2010 2011


Quantity Index 100 84.1 93.2 120.7 121.3

Laspeyres Composite Quantity Index - Motorcycles


130

120

110

100

90

80
Quantity Index
70

60
2007 2008 2009 2010 2011
(g) Interpretation
Sales of motorcycles dropped for two years after 2007 (i.e. by 15.9% in 2008 and
by 6.8% in 2009). Thereafter unit sales, on average per model, increased by 20.7%
in 2010 relative to 2007; but performed at the same level in 2011 compared to 2010.
Exercise 14.28 File: X14.28 - tyre production.xlsx

Data Hillstone Tyre Production Costs and Volumes - Uitenhage Plant

Cost/Tyre 2010-Jan Feb March April May June July Aug Sept Oct Nov Dec
Passenger 210.69 212.47 210.73 218.14 219.22 216.19 225.92 234 229.89 222.76 223.96 200.3
Light truck 376.45 361.7 361.76 363.94 363.62 364.06 376.9 375.4 375.55 375.04 375.59 376.3
Giant truck 1171.1 1109.6 1101.8 1119.7 1127.2 1120.32 1162.8 1181 1157.7 1166.75 1157.7 1148

Output (1000's) 2010-Jan Feb March April May June July Aug Sept Oct Nov Dec
Passenger 78 102 93 81 105 100 117 105 98 110 97 43
Light truck 11 14 13 12 16 16 19 17 14 15 13 6
Giant truck 10 14 13 11 15 15 16 16 16 15 14 5

(a) Cost Relative Index - Passenger Tyres

Cost relative
Jan Feb March April May June July Aug Sept Oct Nov Dec
index series
Passenger tyre 100.0 100.8 100.0 103.5 104.0 102.6 107.2 111.1 109.1 105.7 106.3 95.1

Interpretation
Passenger tyre costs increased steadily throughout the year reaching a peak of 11.1% above January
2010 levels in August. Thereafter costs declined steadily and ended the year 4.9% below the
starting level in January 2010.

(b) PRODUCTION COST ANALYSIS Laspeyres Composite Cost Index (Weighted Aggregates)

Jan Feb March April May June July Aug Sept Oct Nov Dec
Passenger 16433.82 16573 16437 17015 17099 16862.8 17622 18250 17931 17375.3 17469 15623
Light truck 4140.95 3978.7 3979.4 4003.3 3999.8 4004.66 4145.9 4129 4131.1 4125.44 4131.5 4139
Giant truck 11711 11096 11018 11197 11272 11203.2 11628 11812 11577 11667.5 11577 11479
Total Cost 32285.77 31647 31435 32216 32371 32070.7 33396 34192 33639 33168.2 33177 31241

Jan Feb March April May June July Aug Sept Oct Nov Dec
Composite Cost
100.0 98.0 97.4 99.8 100.3 99.3 103.4 105.9 104.2 102.7 102.8 96.8
Index

Composite Cost Index

108.0
106.0
104.0
102.0
100.0
98.0
96.0
94.0
Composite Cost Index
92.0
Jan Feb March April May June July Aug Sept Oct Nov Dec

Interpretation
Composite tyre manufacturing costs declined or remained constant for the first 6 months
until June 2010. Thereafter, costs, on average increased by 6% over July and August
before being brought under control. By December, composite costs were 3.2% below
the beginning of the year levels.
(c) Link Relatives - Costs - Light Truck Tyres

Jan Feb March April May June July Aug Sept Oct Nov Dec
Light truck 376.45 361.7 361.76 363.94 363.62 364.06 376.9 375.4 375.55 375.04 375.59 376.3
Link Relatives 100.0 96.1 100.0 100.6 99.9 100.1 103.5 99.6 100.0 99.9 100.1 100.2

Interpretation
Light truck radial tyres costs have remained almost constant and unchanged throughout
the year with most adjustments not exceeding 0.5%. Only in Feb (decrease of 3.9%) and
July (increase of 3.5%) did costs fluctuate to any degree.

(d) PRODUCTION VOLUME ANALYSIS Laspeyres Quantity Index (Weighted Aggregates)

Jan Feb March April May June July Aug Sept Oct Nov Dec
Passenger 16433.82 21490 19594 17066 22122 21069 24651 22122 20648 23175.9 20437 9060
Light truck 4140.95 5270.3 4893.9 4517.4 6023.2 6023.2 7152.6 6400 5270.3 5646.75 4893.9 2259
Giant truck 11711 16395 15224 12882 17567 17566.5 18738 18738 18738 17566.5 16395 5856
Total Value 32285.77 43156 39712 34465 45712 44658.7 50541 47260 44656 46389.2 41726 17174

Jan Feb March April May June July Aug Sept Oct Nov Dec
Volume Index 100.0 133.7 123.0 106.8 141.6 138.3 156.5 146.4 138.3 143.7 129.2 53.2

Laspeyres Composite Production Output (Volume) Index


180.0
160.0
140.0
120.0
100.0
80.0
60.0
40.0
20.0 Volume Index
0.0
Jan Feb March April May June July Aug Sept Oct Nov Dec

(e) Interpretation
Production volumes of all three makes of tyres showed a steady increase above the January
level by up to nearly 60% over the first half of the year. However output showed a steady
decline in the second half of the year ending at only 53.2% of the beginning of the year levels.
CHAPTER 15

TIME SERIES ANALYSIS

A FORECASTING TOOL

Exercise 15.1 Cross-sectional data is gathered at one point in time;


Time series data is recorded at fixed intervals over time.

Exercise 15.2 Monthly national new car sales;


Daily maximum temperature for Cape Town.

Exercise 15.3 A line graph

Exercise 15.4 Trend; Cycles; Seasonality; Irregular (random)


Seasonality tends to show the most regularity.

Exercise 15.5 The Moving Average method smooths out short-term


fluctuations in a time series to allow the longer-term
underlying trend and cyclical patterns to be revealed.

Exercise 15.6 Yes, averaging occurs over a longer time period (i.e. five periods)
producing a smoother curve.

Exercise 15.7 A seasonal index of 108 means that seasonal influences stimulate
the time series values by 8% above the trend / cyclical level.

Exercise 15.8 A seasonal index of 88 means that seasonal influences depress


the time series values by 12% below the trend / cyclical level.
Exercise 15.9 File: X15.9 - coal tonnage.xlsx

(a) and (b)


Uncentred 4 Centred 4- Centred 5-year
Year Coal Tonnage year moving year Moving Moving
total Average Average
1 118
2 124 470
3 108 484 119.25 120.4
4 120 475 119.875 119.8
5 132 489 120.5 119.4
6 115 517 125.75 127.4
7 122 545 132.75 135.4
8 148 618 145.375 146.6
9 160 697 164.375 163.8
10 188 723 177.5 174.2
11 201 754 184.625 182.8
12 174 744 187.25 186.4
13 191 689 179.125 178
14 178 676 170.625 170
15 146
16 161

(c) and (d)


Line Graphs of Coal mined, 4 and 5 year Moving Averages

250
coal mined (100 000 tonnes)

200

150

Coal Tonnage
100
Centred 4- year Moving
50 Average
Centred 5-year Moving
Average
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

years

(e) Interpretation
The annual tonnage of coal mined in the Limpopo province was constant for the first
7 years after which there was an expansion phase for the next 4 years. Thereafter coal
production has been declining steadily. This could be evidence of a cyclical effect
caused by economic cycles in the demand for coal worldwide.
Overall, a moderate upward cyclical trend.
Exercise 15.10 File: X15.10 - franchise dealers.xlsx

(a) Line Graph - New Franchise Dealers

Line Graph: Trend Line of New Franchise Dealers

60

50
no. of new dealers

40

30

20

10

0
1 2 3 4 5 6 7 8 9 10

time periods

(b) Period (x) Dealers (y) x² xy


1 28 1 28
2 32 4 64
3 43 9 129
4 31 16 124
5 38 25 190
6 47 36 282
7 40 49 280
8 45 64 360
9 55 81 495
n = 10 10 42 100 420
Σ 55 401 385 2372

b1 = ((10*2372)-(55*401))/(10*385-55²) = 2.0182

b0 = (401-2.0182*55)/10 = 29

ŷ= 29 + 2.0182 x x = 1, 2, 3, …, 10

(c) Trend Estimates

x substitute into ŷ ŷ
11 29 + 2.0182 (11) 51.20
12 29 + 2.0182 (12) 53.22
13 29 + 2.0182 (13) 55.24

The number of new franchise dealers are likely to be 51, 53 and 55 in


periods 11, 12 and 13 respectively, based on trend estimates.
Exercise 15.11 File: X15.11 - policy claims.xlsx

(a) Line Graph - Household Policy Claims

Line Graph - Household Policy Claims

90
80
70
60
no. of claims

50
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

quarters

(b) Period (x) Claims (y) x² xy


1 84 1 84
2 53 4 106
3 60 9 180
4 75 16 300
5 81 25 405
6 57 36 342
7 51 49 357
8 73 64 584
9 69 81 621
10 37 100 370
11 40 121 440
12 77 144 924
13 73 169 949
14 46 196 644
15 39 225 585
n = 16 16 63 256 1008
Σ 136 978 1496 7899

b1 = =[(16)(7899) - (136)(978)]/[(16*1496)-(136)2] -1.21765

b0 = =(978-(-1.21765)*136)/16 71.475
1 in Q1 2008
ŷ= 71.475 - 1.21765 x x= 2 in Q2 2008
3 in Q3 2008
Interpretation
There is a downward trend in household policy claims over the past 4 years.
Uncentred 2x4 Centred 4
Unadjusted Adjusted
Time 4 period period period Seasonal
(c) Claims Seasonal Seasonal
Periods Moving Moving Moving ratios
Indexes Indexes
Total Total Average
2008 Q1 84
2008 Q2 53 272
2008 Q3 60 269 541 67.625 88.725 79.688 78.705
2008 Q4 75 273 542 67.75 110.701 121.667 120.166
2009 Q1 81 264 537 67.125 120.670 122.940 121.423
2009 Q2 57 262 526 65.75 86.692 80.702 79.706
2009 Q3 51 250 512 64 79.688
2009 Q4 73 230 480 60 121.667
2010 Q1 69 219 449 56.125 122.940
2010 Q2 37 223 442 55.25 66.968
2010 Q3 40 227 450 56.25 71.111
2010 Q4 77 236 463 57.875 133.045
2011 Q1 73 235 471 58.875 123.992
2011 Q2 46 221 456 57 80.702
2011 Q3 39
2011 Q4 63 Totals 404.996 400.000

Seasonal Indexes Q1 121.4


Q2 79.7
Q3 78.7
Q4 120.2

Interpretation
Household policy claims tend to increase significantly in Quarters 1 and 4 of each year,
by about 20% on average, while there is a significant decline in claims during
Quarters 2 and 3 by about 20% on average.

(d) Seasonally-adjusted trend estimate of houshold policy claims

x = 17 in Quarter 1 2012
Trend estimate ŷ= 71.475 - 1.21765 (17) = 50.77495

x = 18 in Quarter 1 2012
Trend estimate ŷ= 71.475 - 1.21765 (18) = 49.5573

Seasonally-adjusted trend estimate

Q1 2012 ŷ (adj) = 50.77495*1.214 = 61.64


Q2 2012 ŷ (adj) = 49.55742*0.797 = 39.5

Interpretation
The insurance company can expect to receive 62 and 40 (rounded) household policy
claims in the first and second quarters of 2012 respectively.
Exercise 15.12 File: X15.12 - hotel occupancy.xlsx

(a) Line Graph - Monthly Hotel Occupancy Rate (%)

Line Graph - Monthly Hotel Occupancy Rate (%)


100

90
occupancy rate (%)

80

70

60

50

40
1 2 3 4 5 6 7 8 9 10

Months (Sept - June )

(b) Month (x) Rate (y) x² xy


1 74 1 74
2 82 4 164
3 70 9 210
4 90 16 360
5 88 25 440
6 74 36 444
7 64 49 448
8 69 64 552
9 58 81 522
n = 10 10 65 100 650
Σ 55 734 385 3864

b1 = (10*3864 - 55*734)/(10*385 - 55²) = -2.09697

b0 = (734 -(-2.09697)*55)/10 = 84.933

1 in Sept
ŷ= 84.933 - 2.09697 x x= 2 in Oct
3 in Nov
Interpretation
There is a downward trend in hotel occupancy over the past 10 months since September.

(c) Trend estimate Period x ŷ


July 11 61.87%
Aug 12 59.77%

Interpretation
The continued downward trend is reflected in the next 2 month's projections.
Exercise 15.13 File: X15.13 - electricity demand.xlsx

(a), (e), (f) Line Graph - Electricity Demand for a City (Cape Town)

Line Graph - Cape Town Electricity Demand


y = 4.95x + 20.8

180
160
demand (1000 megawatts)

140
120
100
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
quarters (Q1 2008 - Q4 2011)

Interpretation
Electricity demand in Cape Town shows a clear seasonal pattern, peaking in quarter 3
and bottoming out in quarter 4 of each year.

(b) Period (x) Demand (y) x² xy


1 21 1 21
2 42 4 84
3 60 9 180
4 12 16 48
5 35 25 175
6 54 36 324
7 91 49 637
8 14 64 112
9 39 81 351
10 82 100 820
11 136 121 1496
12 28 144 336
13 78 169 1014
14 114 196 1596
15 160 225 2400
n = 16 16 40 256 640
Σ 136 1006 1496 10234

b1 = = (16*10234 - 136*1006)/(16*1496 - 136²) = 4.95

b0 = (1006 -(4.95)*136)/16 = 20.8

1 in Q1 2008
ŷ= 20.8 + 4.95 x x= 2 in Q2 2008
3 in Q3 2008
(c)

Uncentred Centred 4
2x4 period Unadjusted Adjusted
Time 4 period period Seasonal
Demand Moving Seasonal Seasonal
Periods Moving Moving ratios
Total Indexes Indexes
Total Average
2008 Q1 21
2008 Q2 42 135
2008 Q3 60 149 284 35.5 169.01 178.65 175.61
2008 Q4 12 161 310 38.75 30.97 30.97 30.44
2009 Q1 35 192 353 44.125 79.32 79.32 77.97
2009 Q2 54 194 386 48.25 111.92 117.99 115.98
2009 Q3 91 198 392 49 185.71
2009 Q4 14 226 424 53 26.42
2010 Q1 39 271 497 62.125 62.78
2010 Q2 82 285 556 69.5 117.99
2010 Q3 136 324 609 76.125 178.65
2010 Q4 28 356 680 85 32.94
2011 Q1 78 380 736 92 84.78
2011 Q2 114 392 772 96.5 118.13
2011 Q3 160
2011 Q4 40 Totals 406.927 400.000

Seasonal Indexes Q1 77.97


Q2 115.98
Q3 175.61
Q4 30.44

Interpretation
Electricity demand peaks in Q3 by 75% over the trend / cyclical level; and drops
to 70% below the trend /cyclical level during Q4.

(d) Seasonally-adjusted trend estimate of Cape Town's electricity demand

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Q3 2012 19 114.85 175.61 201.69
Q4 2012 20 119.8 30.44 36.47

Interpretation
Electricity demand in Cape Town is likely to peak at 201.69 megawatts in
Q3 of 2012 and bottom out at 36.47 megawatts in Q4 of 2012.
Exercise 15.14 File: X15.14 - hotel turnover.xlsx

Seaonal De-
(a) Actual
Index seasonalised
568 136 417.6
495 112 442.0
252 62 406.5
315 90 350.0
604 136 444.1
544 112 485.7
270 62 435.5
510 90 566.7
662 136 486.8
605 112 540.2
310 62 500.0
535 90 594.4

(b) and (e)


Hotel Industry Quarterly Turnover

700

600
turnover (R millions)

500

400

300

200

100 Actual De-seasonalised


0
1 2 3 4 5 6 7 8 9 10 11 12

quarters (Summer 2008 - Spring 2010)

(c) Trend line estimate ŷ

Period (x) T/over (y) x² xy


1 568 1 568
2 495 4 990
3 252 9 756
4 315 16 1260
5 604 25 3020
6 544 36 3264
7 270 49 1890
8 510 64 4080
9 662 81 5958
10 605 100 6050
11 310 121 3410
n = 12 12 535 144 6420
Σ 78 5670 650 37666
b1 = (12*37666 - 78*5670)/(12*650 - 78²) = 5.6713

b0 = (5670 - (5.6713)*78)/12 = 435.64

1 in Summer 2008
ŷ= 435.64 + 5.6713 x x= 2 in Autumn 2008
3 in Winter 2008

(d) Seasonally-adjusted trend estimate of Hotel Industry Quarterly Turnover

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Summer 2011 13 509.37 136 692.74
Autumn 2011 14 515.04 112 576.84

(f) Trend line estimation using Excel's Chart Wizard

Hotel Industry Quarterly Turnover


y = 5.6713x + 435.64
700

600
turnover (R millions)

500

400

300

200 Actual Linear (Actual)


100
1 2 3 4 5 6 7 8 9 10 11 12

quarters
Exercise 15.15 File: X15.15 - farming equipment.xlsx

Plot of Farming Equipment Sales (2008 - 2011)

Quarterly Farming Implements Sales


y = 0.8544x + 52.05
75
70
65
no. units sold

60
55
50
45
Sales
40
35 Linear (Sales)

30
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

quarters (2008 - 2011)

(a) Seasonal Indexes for Farming Equipment Sales

Uncentred Centred 4
2x4 period Unadjusted Adjusted
4 period period Seasonal
Periods Sales Moving Seasonal Seasonal
Moving Moving Ratios
Total Indexes Indexes
Total Average
2008 Q1 57
2008 Q2 51 214
2008 Q3 50 217 431 53.88 92.81 92.61 92.63
2008 Q4 56 222 439 54.88 102.05 102.95 102.97
2009 Q1 60 225 447 55.88 107.38 107.38 107.40
2009 Q2 56 230 455 56.88 98.46 96.97 97.00
2009 Q3 53 235 465 58.13 91.18
2009 Q4 61 239 474 59.25 102.95
2010 Q1 65 244 483 60.38 107.66
2010 Q2 60 251 495 61.88 96.97
2010 Q3 58 250 501 62.63 92.61
2010 Q4 68 252 502 62.75 108.37
2011 Q1 64 252 504 63.00 101.59
2011 Q2 62 254 506 63.25 98.02
2011 Q3 58
2011 Q4 70 Totals 399.92 400

Seasonal Indexes Summer 107.40


Autumn 97.00
Winter 92.63
Spring 102.97

(b) Seasonal Influences


The influence of seasonal forces on farming equipment sales is modest.
There is a small stimulatory effect during Spring and Summer (no more than 7%) and
a small depressing effect (also no more than 7%) during Autumn and Winter.
(c) Period (x) Sales (y) x² xy
1 57 1 57
2 51 4 102
3 50 9 150
4 56 16 224
5 60 25 300
6 56 36 336
7 53 49 371
8 61 64 488
9 65 81 585
10 60 100 600
11 58 121 638
12 68 144 816
13 64 169 832
14 62 196 868
15 58 225 870
n = 16 16 70 256 1120
Σ 136 949 1496 8357

b1 = (16*8357-136*949)/(16*1496-136²) = 0.8544

b0 = (949 -(0.8544)*136)/16 = 52.05

1 in Summer 2008
ŷ= 52.05 + 0.8544 x x= 2 in Autumn 2008
3 in Winter 2008

(d) Seasonally-adjusted trend estimate of Farming Equipment Sales for 2012

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Summer 17 66.57 107.40 71.50
Autumn 18 67.43 97.00 65.41
Winter 19 68.28 92.63 63.25
Spring 20 69.14 102.97 71.19

Interpretation
The company can expect to sell between 63 and 72 farming implements
each quarter during 2012 with the higher sales expected in Summer and Spring.
Exercise 15.16 File: X15.16 - energy costs.xlsx

(a), (e) and (f) Line Graph of Office Complex Energy Costs (in R100 000)

Office Complex Energy Costs y = 0.0601x + 3.1091

5
4.5
energy cost (R100 000)

4
3.5
3
2.5
2
1.5
1
1 2 3 4 5 6 7 8 9 10 11 12
quarters (2009 - 2011)

(b)

Uncentred 4 2x4 Centred 4


Unadjusted Adjusted
Time period period period Seasonal
Costs Seasonal Seasonal
Periods Moving Moving Moving ratios
Indexes Indexes
Total Total Average
2009 Q1 2.4
2009 Q2 3.8 13.3
2009 Q3 4 13.5 26.80 3.35 119.40 118.27 117.6
2009 Q4 3.1 13.8 27.30 3.4125 90.84 90.49 90.0
2010 Q1 2.6 13.9 27.70 3.4625 75.09 73.41 73.0
2010 Q2 4.1 14.0 27.90 3.4875 117.56 120.21 119.5
2010 Q3 4.1 14.0 28.00 3.5 117.14
2010 Q4 3.2 14.4 28.40 3.55 90.14
2011 Q1 2.6 14.6 29.00 3.625 71.72
2011 Q2 4.5 14.7 29.30 3.6625 122.87
2011 Q3 4.3
2011 Q4 3.3 Totals 402.39 400

Seasonal Indexes Summer 73.0


Autumn 119.5
Winter 117.6
Spring 90.0

Interpretation
Energy costs rise by nearly 20% (19,5% and 17,6%) over the colder months of
Autumn and Winter, and decline by between 10% (in Spring) and almost 30%
(27%) in Summer.
(c) Period (x) Cost (y) x² xy
1 2.4 1 2.4
2 3.8 4 7.6
3 4 9 12
4 3.1 16 12.4
5 2.6 25 13
6 4.1 36 24.6
7 4.1 49 28.7
8 3.2 64 25.6
9 2.6 81 23.4
10 4.5 100 45
11 4.3 121 47.3
n = 12 12 3.3 144 39.6
Σ 78 42 650 281.6

2
b1 = [(12*281.6)-(78*42)]/[12*650-78 ] = 0.0601

b0 = (42 - 0.0602*78)/12 = 3.1091

1 in Summer 2009
ŷ= 3.1091 + 0.0601 x x= 2 in Autumn 2009
3 in Winter 2009

(d) Seasonally-adjusted trend estimate of Office Complex Energy Costs for 2012

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Summer 13 3.89 72.97 2.84
Autumn 14 3.95 119.50 4.72
Winter 15 4.01 117.57 4.71
Spring 16 4.07 89.95 3.66

Interpretation
The office complex manager must budget between R284 000 and R472 000 each
quarter during 2012 with higher costs expected in the Autumn and Winter periods.
Exercise 15.17 File: X15.17 - business registrations.xlsx

(a) Line Graph of New Business Registrations (2007 - 2011)

Line Graph - New Business Registrations

2500
no. of new registrations

2000

1500

1000

500

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

quarters (2007 - 2011)

(b) and (c) 4-Period Moving Average and Quarterly Seasonal Indexes

(b) (c)
2x4 Centred 4
Uncentred 4 Unadjusted Adjusted
Time New period period Seasonal
period Moving Seasonal Seasonal
Periods Registrations Moving Moving ratios
Total Indexes Indexes
Total Average
2007 Q1 1005
2007 Q2 1222 4724
2007 Q3 1298 4892 9616 1202.00 107.99 106.526 106.41
2007 Q4 1199 5041 9933 1241.63 96.57 98.514 98.41
2008 Q1 1173 5199 10240 1280.00 91.64 91.820 91.72
2008 Q2 1371 5376 10575 1321.88 103.72 103.568 103.46
2008 Q3 1456 5517 10893 1361.63 106.93
2008 Q4 1376 5677 11194 1399.25 98.34
2009 Q1 1314 5826 11503 1437.88 91.38
2009 Q2 1531 5980 11806 1475.75 103.74
2009 Q3 1605 6125 12105 1513.13 106.07
2009 Q4 1530 6265 12390 1548.75 98.79
2010 Q1 1459 6422 12687 1585.88 92.00
2010 Q2 1671 6569 12991 1623.88 102.90
2010 Q3 1762 6714 13283 1660.38 106.12
2010 Q4 1677 6880 13594 1699.25 98.69
2011 Q1 1604 7034 13914 1739.25 92.22
2011 Q2 1837 7176 14210 1776.25 103.42
2011 Q3 1916
2011 Q4 1819 Totals 400.429 400

Seasonal Indexes Q1 91.72


Q2 103.46
Q3 106.41
Q4 98.41
(d) Interpretation of Seasonal Influences on new business registrations
New business registrations show a modest seasonal pattern, ranging between
8,3% below the annual average in quarter 1 to 6.4% above the annual average
in quarter 3.

(b) 4-Period Moving Average Line Graph and Original Data Line Graph

Line Graph - New Business Registrations


2500

2000
no. of new businesses

1500

1000

New Registrations
500
Centred 4 period Moving
Average
0

quarters (2007 - 2011)

(e) Least Squares trend line (using Add Trendline in Excel )


1 in Q1 2007
ŷ= 1078.2 + 39.338 x x= 2 in Q2 2007
3 in Q3 2007
(f) Seasonally-adjusted Trend estimates

Adjusted
Trend
Periods y Seasonal ŷ (adj)
Estimate ŷ
Indexes
2007 Q1 1005 91.72 1117.5 1025.0
2007 Q2 1222 103.46 1156.9 1196.9
2007 Q3 1298 106.41 1196.2 1272.9
2007 Q4 1199 98.41 1235.6 1215.9
2008 Q1 1173 91.72 1274.9 1169.3
2008 Q2 1371 103.46 1314.2 1359.7
2008 Q3 1456 106.41 1353.6 1440.3
2008 Q4 1376 98.41 1392.9 1370.8
2009 Q1 1314 91.72 1432.2 1313.7
2009 Q2 1531 103.46 1471.6 1522.5
2009 Q3 1605 106.41 1510.9 1607.8
2009 Q4 1530 98.41 1550.3 1525.6
2010 Q1 1459 91.72 1589.6 1458.0
2010 Q2 1671 103.46 1628.9 1685.3
2010 Q3 1762 106.41 1668.3 1775.2
2010 Q4 1677 98.41 1707.6 1680.5
2011 Q1 1604 91.72 1746.9 1602.3
2011 Q2 1837 103.46 1786.3 1848.1
2011 Q3 1916 106.41 1825.6 1942.6
2011 Q4 1819 98.41 1865.0 1835.3

(g) Plot of Actual vs Seasonally adjusted Trend estimates (New Business Registrations)

New Business Registrations (Actual vs Trend Adjusted)

2000
no. of new registrations

1800
1600
1400
1200
1000 y ŷ (adj)
800
2007 Q1

2007 Q2

2007 Q3

2007 Q4

2008 Q1

2008 Q2

2008 Q3

2008 Q4

2009 Q1

2009 Q2

2009 Q3

2009 Q4

2010 Q1

2010 Q2

2010 Q3

2010 Q4

2011 Q1

2011 Q2

2011 Q3

2011 Q4

quarters (2007 - 2011)

Comment
The seasonally adjusted trend estimates track the actual number of new business
registrations very closely. It is a good fitting graph.
(h) Seasonally-adjusted trend estimate of New Business Registrations (2012)

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Q1 2012 21 1904.3 91.72 1746.6
Q2 2012 22 1943.6 103.46 2010.9
Q3 2012 23 1983.0 106.41 2110.1
Q4 2012 24 2022.3 98.41 1990.2
Exercise 15.18 File: X15.18 - engineering sales.xlsx

Total Sales (Estimate for next year)

Seasonally
Seasonal
Quarter Trend ŷ Adjusted Trend
Index
Estimate ŷ(adj)
1 95 12 11.4
2 115 12 13.8
3 110 12 13.2
4 80 12 9.6
48 48
Exercise 15.19 File: X15.19 - Table Mountain.xlsx

(a) and (b) Period Visitors S Index De-Seasonlised


Wi 2009 18.1 62 29.19
Sp 2009 26.4 89 29.66
Su 2009 41.2 162 25.43
Au 2009 31.6 87 36.32
Wi 2010 22.4 62 36.13
Sp 2010 33.2 89 37.30
Su 2010 44.8 162 27.65
Au 2010 32.5 87 37.36

Table Mountain Visitors


(Actual vs De-Seasonalised)
50
45
no. of visitors (1000's)

40
35
30
25
20
15
10 Visitors
5 De-Seasonlised
0
Wi 2009 Sp 2009 Su 2009 Au 2009 Wi 2010 Sp 2010 Su 2010 Au 2010

Quarters (2009 - 2010)

Comment
The trend (after de-seasonalising the quarterly data) is only marginally upwards.

(c) Table Mountain Visitors (per quarter for 2011)

Seasonally
Quarter Seasonal Index Trend ŷ Adjusted Trend
Estimate ŷ(adj)

Winter 62 37.5 23.25


Spring 89 37.5 33.375
Summer 162 37.5 60.75
Autumn 87 37.5 32.625
150 150
Exercise 15.20 File: X15.20 - gross domestic product.xlsx

(a) Year GDP (Millions)


2001 45
2002 47
2003 61
2004 64
2005 72
2006 74
2007 84
2008 81
2009 93
2010 90
2011 98

African Country Gross


Domestic Product y = 5.2636x + 41.964
120

100
GDP (in R100 millions)

80

60

40
GDP (Millions)
20 Linear (GDP (Millions))

0
1 2 3 4 5 6 7 8 9 10 11

years (2001 - 2011)

1 in 2001
(b) Trend line (see graph in (i)) ŷ = 41.964 + 5.2636 x x = 2 in 2002
3 in 2003

(c) Expected GDP for 2012 and 2013 (in R100 million)

Year Year (x ) Trend ŷ Estimate (ŷ)


2012 12 41.964 + 5.2636 (12) 105.1
2013 13 41.964 + 5.2636 (13) 110.4
Exercise 15.21 File: X15.21 - pelagic fish.xlsx

(a) and (c) (a) (c)


Unadj
3 Period Seasonal Seasonal
Year Months Pelagic fish Seasonal
Moving Ave ratio Index
Index
2007 Jan - Apr 44
May - Aug 36 38.00 94.74 96.85 97.79
Sept - Dec 34 38.33 88.70 88.70 89.56
2008 Jan - Apr 45 40.33 111.57 111.57 112.65
May - Aug 42 40.33 104.13
Sept - Dec 34 38.00 89.47
2009 Jan - Apr 38 34.67 109.62
May - Aug 32 32.33 98.97
Sept - Dec 27 33.00 81.82
2010 Jan - Apr 40 32.67 122.45
May - Aug 31 33.00 93.94
Sept - Dec 28 Totals 297.12 300

Seasonal
(c) Periods
Indexes
Jan - Apr 112.65
May - Aug 97.79
Sept - Dec 89.56

Pelagic fish catches are:


- significantly higher (12.65%) than the trend /cyclical volumes in the Jan - April period.
- only marginally lower (2.21%) than the trend /cyclical volumes in the May - Aug period.
- more than 10% lower than the trend /cyclical volumes in the Sept - Dec period.

(b)

Pelagic Fish Catches


(Actual vs 3-Period Moving Average)
50
45
40
35
no. of tonnes

30
25
20
15
10 Pelagic fish
5
3 Period Moving Ave
0
1 2 3 4 5 6 7 8 9 10 11 12

2003 - 2006 (4-monthly)

Comment
The trend in pelagic fish catches over the past 4 years is decidedly downwards.
(d) Period (x) Catch (y) x² xy
1 44 1 44
2 36 4 72
3 34 9 102
4 45 16 180
5 42 25 210
6 34 36 204
7 38 49 266
8 32 64 256
9 27 81 243
10 40 100 400
11 31 121 341
n = 12 12 28 144 336
Σ 78 431 650 2654

b1 = (12*2654 - 78*431)/(12*650 - 78²) = -1.0315

b0 = (431 - (-1.031469)*78)/12 = 42.621

1 in Jan - Apr 2007


ŷ= 42.621 - 1.0315 x x= 2 in May - Aug 2007
3 in Sept - Dec 2007

(e) Seasonally-adjusted trend estimates of Pelagic Fish Catches for 2011

Seasonally
Period Seasonal
x Trend ŷ adjusted
2011 Index
Trend
Jan - Apr 13 29.21 112.65 32.91
May - Aug 14 28.18 97.79 27.56
Sept - Dec 15 27.15 89.56 24.32
Exercise 15.22 File: X15.22 - share price.xlsx

Month Period Share price Trend Price calc Trend Price


January 1 90 = 89.428 - 2.7143(1) 86.7
February 2 82 = 89.428 - 2.7143(2) 84.0
March 3 78 = 89.428 - 2.7143(3) 81.3
April 4 80 = 89.428 - 2.7143(4) 78.6
May 5 74 = 89.428 - 2.7143(5) 75.9
June 6 76 = 89.428 - 2.7143(6) 73.1
July 7 70 = 89.428 - 2.7143(7) 70.4
Aug 8 = 89.428 - 2.7143(8) 67.7
Sept 9 = 89.428 - 2.7143(9) 65.0
Oct 10 = 89.428 - 2.7143(10) 62.3
Nov 11 = 89.428 - 2.7143(11) 59.6

Trend equation (derived from Excel 's Add Trendline)


1 in January
ŷ = 89.429 - 2.7143 x x = 2 in February
3 in March

Threshhold selling price 90 - 1/3 (90) = 60c

Share Price Trend


y = 89.429 - 2.7143 x
100

90
price (cents)

80

70

60
Share price
50
Trend Price
40
1 2 3 4 5 6 7 8 9 10 11

months (Jan - Nov)

Conclusion
The trend estimate of the share price is likely to fall below 60c (the selling level)
by November of the same year if the downward trend continues uninterrupted.
Exercise 15.23 File: X15.23 - Addo park.xlsx

Seaonal De-
(a) Season Actual
Index seasonalised
Su 2008 196 112 175.0
Au 2008 147 94 156.4
Wi 2008 124 88 140.9
Sp 2008 177 106 167.0
Su 2009 199 112 177.7
Au 2009 152 94 161.7
Wi 2009 132 88 150.0
Sp 2009 190 106 179.2
Su 2010 214 112 191.1
Au 2010 163 94 173.4
Wi 2010 145 88 164.8
Sp 2010 198 106 186.8

(b) and (d) Line Plots of Actual and De-Seasonalised Visitor Numbers (by Season)

Addo National Park Visitors


(Actual vs De-Seasonalised)
250
no. of visitors (in 1000s)

200

150

100
Actual
50
De-seasonalised
0
Su Au Wi Sp Su Au Wi Sp Su Au Wi Sp
2008 2008 2008 2008 2009 2009 2009 2009 2010 2010 2010 2010

quarters (2004 - 2006)

(c) Conclusion
There is a very slight upward trend in visitors to the Addo National Park. There is
almost no growth in visitors over the past 3 years.
Exercise 15.24 File: X15.24 - healthcare claims.xlsx

Value of Healthcare Claims (in R millions)

(a) Seasonal Indexes for Healthcare Claims

Uncentred Centred 4
2x4 period Unadjusted Adjusted
4 period period Seasonal
Periods Claims Moving Seasonal Seasonal
Moving Moving Ratios
Total Indexes Indexes
Total Average
2007 Q1 11.8
2007 Q2 13.2 60.5
2007 Q3 19.1 59.6 120.1 15.0125 127.23 139.67 136.28
2007 Q4 16.4 58.8 118.4 14.8 110.81 106.75 104.15
2008 Q1 10.9 62.1 120.9 15.1125 72.13 72.13 70.37
2008 Q2 12.4 63.5 125.6 15.7 78.98 91.41 89.19
2008 Q3 22.4 64.8 128.3 16.0375 139.67
2008 Q4 17.8 68.6 133.4 16.675 106.75 409.96 400
2009 Q1 12.2 70.3 138.9 17.3625 70.27
2009 Q2 16.2 67.1 137.4 17.175 94.32
2009 Q3 24.1 67.7 134.8 16.85 143.03
2009 Q4 14.6 66 133.7 16.7125 87.36
2010 Q1 12.8 62.7 128.7 16.0875 79.56
2010 Q2 14.5 64.2 126.9 15.8625 91.41
2010 Q3 20.8
2010 Q4 16.1

Seasonal Indexes Q1 70.37


Q2 89.19
Q3 136.28
Q4 104.15

(b) Period (x) Claims (y) x² xy


1 11.8 1 11.8
2 13.2 4 26.4
3 19.1 9 57.3
4 16.4 16 65.6
5 10.9 25 54.5
6 12.4 36 74.4
7 22.4 49 156.8
8 17.8 64 142.4
9 12.2 81 109.8
10 16.2 100 162
11 24.1 121 265.1
12 14.6 144 175.2
13 12.8 169 166.4
14 14.5 196 203
15 20.8 225 312
n = 16 16 16.1 256 257.6
Σ 136 255.3 1496 2240.3
b1 = (16*2240.3 - 136*255.3)/(16*1496 - 136²) = 0.2066

b0 = (255.3 -(0.206618)*136)/16 = 14.2

1 in Q1 2007
ŷ= 14.2 + 0.2066 x x= 2 in Q2 2007
3 in Q3 2007

(c) Seasonally-adjusted trend estimate of Healthcare Claims for 2011

Seasonally
Seasonal
Period x Trend ŷ adjusted
Index
Trend
Q1 2011 17 17.71 70.37 12.46
Q2 2011 18 17.92 89.19 15.98
Q3 2011 19 18.13 136.28 24.71
Q4 2011 20 18.33 104.15 19.09

Interpretation
Healthcare claims are expected to rise during 2011 from a low of
R12,46 (mill) in Q1 to R24,71 (mill) in Q3.

(d) Line Plots of Healthcare Claims (Actual vs Seasonally-adjusted trend estimates)

Periods Claims S Index Time Trend (ŷ) Trend (ŷ-adj)


2007 Q1 11.8 70.37 1 14.41 10.14
2007 Q2 13.2 89.19 2 14.61 13.03
2007 Q3 19.1 136.28 3 14.82 20.20
2007 Q4 16.4 104.15 4 15.03 15.65
2008 Q1 10.9 70.37 5 15.23 10.72
2008 Q2 12.4 89.19 6 15.44 13.77
2008 Q3 22.4 136.28 7 15.65 21.32
2008 Q4 17.8 104.15 8 15.85 16.51
2009 Q1 12.2 70.37 9 16.06 11.30
2009 Q2 16.2 89.19 10 16.27 14.51
2009 Q3 24.1 136.28 11 16.47 22.45
2009 Q4 14.6 104.15 12 16.68 17.37
2010 Q1 12.8 70.37 13 16.89 11.88
2010 Q2 14.5 89.19 14 17.09 15.24
2010 Q3 20.8 136.28 15 17.30 23.58
2010 Q4 16.1 104.15 16 17.51 18.23
2011 Q1 70.37 17 17.71 12.46
2011 Q2 89.19 18 17.92 15.98
2011 Q3 136.28 19 18.13 24.70
2011 Q4 104.15 20 18.33 19.09
Healthcare Claims
(Actual vs Seasonally-adjusted Trend Estimates)

30

Claims (in R millions)


25

20

15

10

5 Claims Trend (ŷ-adj)

quarters (2007 - 2010)


Exercise 15.25 File: X15.25 - financial advertising.xlsx

(a) Year Time (x) Exp (y) x² xy


2005 1 9.6 1 9.6
2006 2 11.8 4 23.6
2007 3 12 9 36
2008 4 13.6 16 54.4
2009 5 14.1 25 70.5
2010 6 15 36 90
2011 7 17.8 49 124.6
Sum 28 93.9 140 408.7
n =7

b1 = (7*408.7-28*93.9)/(7*140-282) = 1.1821

b0 = (93.9-1.182143*28)/7 8.6857

1 in 2005
ŷ = 8.6857 + 1.1821 x x= 2 in 2006
3 in 2007

(b) ŷ (2012) = 8.6857 + 1.1821(8) = 18.143

(c)
Financial Services Sector
Advertising Expenditure y = 1.1821x + 8.6857

20
Adspend (in R10 millions)

18

16

14

12

10

8
1 2 3 4 5 6 7

years (2005 - 2011)

(d) Trend line equation (using Excel 's Add Trendline function)

1 in 2005
ŷ = 8.6857 + 1.1821 x x= 2 in 2006
3 in 2007
(e) Year Time (x) Actual (y) Trend (ŷ)
2005 1 9.6 9.87
2006 2 11.8 11.05
2007 3 12 12.23
2008 4 13.6 13.41
2009 5 14.1 14.60
2010 6 15 15.78
2011 7 17.8 16.96
2012 8 18.14

Financial Services Adspend

20
adspend (in R10 million)

18

16

14

12
Actual (y)
10 Trend (ŷ)
8
1 2 3 4 5 6 7 8

years (2005 - 2011)


Exercise 15.26 File: X15.26 - policy surrenders.xlsx

(a) Line Graph of Surrendered Endowment Policies (2008 - 2010)

Surrendered Endowment Policies

240
220
200
no. of policies

180
160
140
120 Policies
100
2008 2008 2008 2008 2009 2009 2009 2009 2010 2010 2010 2010
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4

quarters (2008 - 2010)

Interpretation
There is a distinct moderate downward trend in the number of surrendered policies
over the period 2008 - 2010. This reduction could be due to the improved client
communication strategy adopted by the company in recent years.

(b)

Uncentred Centred 4
2x4 period Unadjusted Adjusted
Time 4 period period Seasonal
Policies Moving Seasonal Seasonal
Periods Moving Moving ratios
Total Indexes Indexes
Total Average
2008 Q1 212
2008 Q2 186 795
2008 Q3 192 769 1564 195.5 98.21 97.84 97.99
2008 Q4 205 748 1517 189.625 108.11 107.47 107.62
2009 Q1 186 725 1473 184.125 101.02 100.62 100.77
2009 Q2 165 702 1427 178.375 92.50 93.49 93.62
2009 Q3 169 685 1387 173.375 97.48
2009 Q4 182 678 1363 170.375 106.82
2010 Q1 169 671 1349 168.625 100.22
2010 Q2 158 667 1338 167.25 94.47
2010 Q3 162
2010 Q4 178 Totals 399.41 400

Seasonal Indexes Q1 100.77


Q2 93.62
Q3 97.99
Q4 107.62

Interpretation
Endowment policy surrenders are highest in Q4 by 7,62% over the quarterly average;
and lowest in Q2 by 6,38% below the quarterly average. There is very little - to none -
seasonal impact on policy surrenders in Q1 and only 2,01% below the quarterly
average in Q3.
(c) Period (x) Cost (y) x² xy
1 212 1 212
2 186 4 372
3 192 9 576
4 205 16 820
5 186 25 930
6 165 36 990
7 169 49 1183
8 182 64 1456
9 169 81 1521
10 158 100 1580
11 162 121 1782
n = 12 12 178 144 2136
Σ 78 2164 650 13558

b1 = (12*13558 - 78*2164)/(12*650-78²) = -3.55245

b0 = (2164 -(-3.552448)*78)/12 = 203.42

1 in Q1 2008
ŷ= 203.42 - 3.5524 x x= 2 in Q2 2008
3 in Q3 2008

(d) Seasonally-adjusted trend estimate of no. of Surrendered Endowment Policies in2011

Seasonally
Seasonal
Period x Trend ŷ adjusted (Rounded)
Index
Trend
Q1 2011 13 157.24 100.77 158.45 158
Q2 2011 14 153.69 93.62 143.88 144
Q3 2011 15 150.13 97.99 147.12 147
Q4 2011 16 146.58 107.62 157.75 158
Exercise 15.27 File: X15.27 - company liquidations.xlsx

(a)
Period Liquidations 3-per M A 5-per M A
1 246
2 243 252.7
3 269 289.7 255.6
4 357 263.0 237.2
5 163 224.7 210.4
6 154 142.0 189.0
7 109 141.7 162.0
8 162 164.3 184.0
9 222 219.0 210.0
10 273 259.7 249.2
11 284 287.3 275.4
12 305 294.0 300.6
13 293 315.3 330.6
14 348 354.7 332.0
15 423 354.0 335.0
16 291 344.7 327.0
17 320 288.0 304.2
18 253 269.0 252.0
19 234 216.3 241.8
20 162 212.0 237.4
21 240 233.3 239.6
22 298 267.3 243.4
23 264 271.7 269.6
24 253 270.0 282.0
25 293 282.7 260.0
26 302 261.0
27 188

(b)

Company Liquidations
(Actual vs 3 and 5 period Moving Averages)
450
400
350
no. of liquidations

300
250
200
150
100
50
Liquidations 3-per M A 5-per M A
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27

periods

(c) Interpretation
The level of business liquidations shows no actual upward / downward trend over the
past 27 periods. There is a distinct cyclical pattern wih the longest lasting from
period 7 to period 20.
Exercise 15.28 File: X15.28 - passenger tyres.xlsx

Passenger Tyre Sales (Y)

(a) Quarterly Seasonal Ratios and Trend Line Equation

Centred 4
Uncentred 4 Unadjusted Adjusted
Time 2x4 period period Seasonal
Tyre Sales period Seasonal Seasonal
Periods Moving Total Moving ratios
Moving Total Indexes Indexes
Average
2005 Q1 64876
2005 Q2 58987 240829
2005 Q3 54621 244699 485528 60691.0 90.00 90.64 90.72
2005 Q4 62345 252285 496984 62123.0 100.36 100.81 100.91
2006 Q1 68746 258591 510876 63859.5 107.65 107.65 107.76
2006 Q2 66573 267480 526071 65758.9 101.24 100.51 100.61
2006 Q3 60927 277522 545002 68125.3 89.43
2006 Q4 71234 282186 559708 69963.5 101.82
2007 Q1 78788 289357 571543 71442.9 110.28
2007 Q2 71237 292567 581924 72740.5 97.93
2007 Q3 68098 291438 584005 73000.6 93.28
2007 Q4 74444 296653 588091 73511.4 101.27 399.62 400.00
2008 Q1 77659 302011 598664 74833.0 103.78
2008 Q2 76452 306475 608486 76060.8 100.51
2008 Q3 73456 313379 619854 77481.8 94.80
2008 Q4 78908 318170 631549 78943.6 99.95
2009 Q1 84563 319592 637762 79720.3 106.07
2009 Q2 81243 327440 647032 80879.0 100.45
2009 Q3 74878 334433 661873 82734.1 90.50
2009 Q4 86756 338248 672681 84085.1 103.18
2010 Q1 91556 340405 678653 84831.6 107.93
2010 Q2 85058 333794 674199 84274.9 100.93
2010 Q3 77035 345161 678955 84869.4 90.77
2010 Q4 80145 356559 701720 87715.0 91.37
2011 Q1 102923
2011 Q2 96456

(a) Seasonal Indexes Q1 107.76


Q2 100.61
Q3 90.72
Q4 100.91

Trend line Use Excel 's Add Trendline function


Passenger Car Tyre Sales y = 1302x + 58114

120000

units sold (in 1000's)


100000

80000

60000

40000 Tyre Sales

20000 Linear (Tyre Sales)

quarters (2000 - 2006)

1 in Q1 2005
(b) ŷ = 58114 + 1302 x x = 2 in Q2 2005
3 in Q3 2005

(b) Seasonally-adjusted Trend estimates of passenger car tyre sales (in units) for 2012 / 2013.

Seasonally
Time Seasonal Adjusted
Time Trend (ŷ)
Periods Indices Trend
Estimate
2012 Q1 29 95872 107.76 103312
2012 Q2 30 97174 100.61 97767
2012 Q3 31 98476 90.72 89337
2012 Q4 32 99778 100.91 100686
2013 Q1 33 101080 107.76 108924
2013 Q2 34 102382 100.61 103007
2013 Q3 35 103684 90.72 94062
2013 Q4 36 104986 100.91 105941

(c) Interpretation
The pattern of passenger car tyre sales is very stable.
The trend is linear and upward, and seasonal variations are consistent over time.
Hence Hillstone management could have high confidence in the estimates.
Exercise 15.29 File: X15.29 - outpatient attendances.xlsx

Outpatients Attendances - Butterworth Clinic

(a) Quarterly Seasonal Ratios and Trend Line Equation

Uncentred Centred 4
2x4 period Unadjusted Adjusted
Time 4 period period Seasonal
Visits Moving Seasonal Seasonal
Periods Moving Moving ratios
Total Indexes Indexes
Total Average
2006 Q1 12767
2006 Q2 16389 64041
2006 Q3 19105 65472 129513 16189.1 118.01 117.31 117.46
2006 Q4 15780 68951 134423 16802.9 93.91 92.62 92.75
2007 Q1 14198 70745 139696 17462.0 81.31 78.99 79.10
2007 Q2 19868 73269 144014 18001.8 110.37 110.53 110.69
2007 Q3 20899 73712 146981 18372.6 113.75
2007 Q4 18304 74048 147760 18470.0 99.10
2008 Q1 14641 74227 148275 18534.4 78.99
2008 Q2 20204 72000 146227 18278.4 110.53
2008 Q3 21078 71434 143434 17929.3 117.56
2008 Q4 16077 72489 143923 17990.4 89.36 399.46 400.00
2009 Q1 14075 72378 144867 18108.4 77.73
2009 Q2 21259 72484 144862 18107.8 117.40
2009 Q3 20967 70821 143305 17913.1 117.05
2009 Q4 16183 71386 142207 17775.9 91.04
2010 Q1 12412 72569 143955 17994.4 68.98
2010 Q2 21824 74365 146934 18366.8 118.82
2010 Q3 22150 79370 153735 19216.9 115.26
2010 Q4 17979 78114 157484 19685.5 91.33
2011 Q1 17417 80274 158388 19798.5 87.97
2011 Q2 20568 85413 165687 20710.9 99.31
2011 Q3 24310 67996 153409 19176.1 126.77
2011 Q4 23118 47428 115424 14428.0 160.23

Seasonal Indexes Q1 79.10


Q2 110.69
Q3 117.46
Q4 92.75

Trend line Use Excel 's Add Trendline function


Outpatient Visits
y = 224.66x + 15591

30000

25000
no. of patients
20000

15000

10000
Visits
5000
Linear (Visits)
0
1 3 5 7 9 11 13 15 17 19 21 23

quarters (2006 - 2011)

1 in Q1 2006
ŷ = 15591 + 224.66 x x = 2 in Q2 2006
3 in Q3 2006

(b) Seasonally-adjusted Trend estimates of Outpatient Visits (Butterworth)


for 2012 / 2013 (first half).

Seasonally
Time Seasonal Adjusted
Time Trend (ŷ)
Periods Indices Trend
Estimate
2012 Q1 25 21207.5 79.10 16775
2012 Q2 26 21432.2 110.69 23723
2012 Q3 27 21656.8 117.46 25438
2012 Q4 28 21881.5 92.75 20295
2013 Q1 29 22106.1 79.10 17486
2013 Q2 30 22330.8 110.69 24718

(c) Interpretation
The pattern of outpatient attendances at the Butterworth Clinic is very stable.
It shows a steady upward trend with highly consistent seasonal variations.
Demand increases in the winter months ((Q3) and is lowest in the summer months (Q1).
Exercise 15.30 File: X15.30 - construction absenteeism.xlsx

Construction Absenteeism

(a) Quarterly Seasonal Ratios and Trend Line Equation

Uncentred 4 2x4 Centred 4


Unadjusted Adjusted
Time period period period Seasonal
Days_lost Seasonal Seasonal
Periods Moving Moving Moving ratios
Indexes Indexes
Total Total Average
2006 Q1 933
2006 Q2 865 3584
2006 Q3 922 3618 7202 900.3 102.42 102.74 102.88
2006 Q4 864 3689 7307 913.4 94.59 95.80 95.93
2007 Q1 967 3698 7387 923.4 104.72 104.07 104.21
2007 Q2 936 3736 7434 929.3 100.73 96.85 96.98
2007 Q3 931 3661 7397 924.6 100.69
2007 Q4 902 3570 7231 903.9 99.79
2008 Q1 892 3546 7116 889.5 100.28
2008 Q2 845 3445 6991 873.9 96.70
2008 Q3 907 3368 6813 851.6 106.50
2008 Q4 801 3238 6606 825.8 97.00 399.45 400.00
2009 Q1 815 3110 6348 793.5 102.71
2009 Q2 715 3020 6130 766.3 93.31
2009 Q3 779 3027 6047 755.9 103.06
2009 Q4 711 3168 6195 774.4 91.82
2010 Q1 822 3151 6319 789.9 104.07
2010 Q2 856 3162 6313 789.1 108.47
2010 Q3 762 3125 6287 785.9 96.96
2010 Q4 722 2984 6109 763.6 94.55
2011 Q1 785 2962 5946 743.3 105.62
2011 Q2 715 2944 5906 738.3 96.85
2011 Q3 740 2159 5103 637.9 116.01
2011 Q4 704 1444 3603 450.4 156.31

Seasonal Indexes Q1 104.21


Q2 96.98
Q3 102.88
Q4 95.93

Trend line Use Excel 's Add Trendline function

Construction Industry Days Lost due to Absenteeism


y = -9.917x + 952.75
1000

900
no. of days lsot

800

700

600
Days_lost
500
Linear (Days_lost)
400
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

quarters (2006 - 2011)


1 in Q1 2006
ŷ = 952.75 - 9.917 x x = 2 in Q2 2006
3 in Q3 2006

(b) Seasonally-adjusted Trend estimates of Days Lost in Construction Industry for 2012.

Seasonally
Time Seasonal Adjusted
Time Trend (ŷ)
Periods Indices Trend
Estimate
2012 Q1 25 704.8 104.21 734
2012 Q2 26 694.9 96.98 674
2012 Q3 27 685.0 102.88 705
2012 Q4 28 675.1 95.93 648

(c) Interpretation
The pattern of days lost due to absenteeism in the Construction industry shows
a distinct downward trend but with inconsistent seasonal variations.
CHAPTER 16

FINANCIAL CALCULATIONS

INTEREST, ANNUITIES and NPV

Exercise 16.1 Simple interest


Interest is computed on the original lump sum for each period.
Compound interest
For each period, interest is computed on the original lump sum
plus all accummulated interest of the preceeding periods.

Exercise 16.2 No - a compounded amount will earn more interest than a


simple interest investment.

Exercise 16.3 Yes - compounding quarterly will result in interest being


capitalised sooner - and therefore earning more interest than
an annual compounded investment.

Exercise 16.4 Nominal interest rate is the quoted per annum interest rate
Effective interest rate is the actual interest rate achieved
when interest is compounded more than once per year.

Exercise 16.5 Annuity - an annuity is when a constant sum of money is paid


(or received) at regular intervals over a period of time.

Exercise 16.6 Ordinary annuity - regular payments begin the first period of the annuity term
Deferred annuity - regular payments begin only at some future period into
the term of the annuity.

Exercise 16.7 Ordinary annuity certain - the series of regular payments take place at the
end of each payment period.

Ordinary annuity due - the series of regular payments take place at the
beginning of each payment period.

Exercise 16.8 NPV is the term used to convert all cash inflows (and outflows) over time
into present value terms by dividing by the annual rate of interest
It represents future cash flows in current terms.
Exercise 16.9

(a) Fv = 15000*(1+0.08*5) = R21 000.00

(b) Fv = 15000*(1+0.08)5 = R22 039.92

(c) Fv = 15000*(1+0.04)10 = R22 203.66


Exercise 16.10

(a) Pv = 150000/(1+0.12)² = R119 579.10

(b) Pv = 150000/(1+0.06)4 = R118 814

(c) Pv = 150000/(1+0.01)24 = R118 134.90


Exercise 16.11

(a) n = (3/1-1)/0.16 = 12.5 years

(b) n = log(3/1)/log(1+0.16) = 7.402 years

(c) n = log(3/1)/log(1+0.04) = 28.011 quarters


(or 7.003 years)
Exercise 16.12

(a) Pv = 10525/(1+0.14/12*30) = R 7,796.30

(b) Pv = 10525/(1+0.14)2.5 R 7,585.08


Exercise 16.13

(a) Effective Rate =(1+0.15/12)12-1 = 16.0755%

(b) =EFFECT(0.15,12) 16.0755%


Exercise 16.14

(a) Effective Rate =(1+0.09/4)4-1 = 9.3083%

(b) =EFFECT(0.09,4) 9.3083%


Exercise 16.15

n = log(58890/25000)/log(1+0.11/2) = 16.00267 half years

n = 8.0013 years
Exercise 16.16

Part 1 - First 3 months


Fv = 2000*(1+0.10/2)0.5 = R 2,049.39

Part 2 - Remaining 21 months


Fv = 2049.39*(1+0.12/12)21 = R 2,525.65

The value of the investment after 2 years is R 2,525.65


Exercise 16.17

Quarterly rate (i ) = (10200/7500)1/(3*4) - 1 = 0.02595% per quarter

Annual rate = (0.02595*4) = 10.3819% p.a.


Exercise 16.18

Monthly rate (i ) = (8000/5000)1/(4*12) - 1 = 0.00984% per month

Annual rate = (0.00984*12) = 11.8078% p.a.


Exercise 16.19

Pv = 25000/(1+0.09/4)11 = R 19,572.37

The investor must deposit R19 572.37 today.


Exercise 16.20

n = log(30000/21353.4)/log(1+0.12) = 3 years
Exercise 16.21

Let Pv = R1 and Fv = R2

Quarterly rate (i ) = (2/1)1/(7*4) - 1 = 0.025064

Annual rate (%) = 0.025064*4 = 10.0257% p.a.


Exercise 16.22

Ordinary Annuity Certain

(a) Fv = 1600*((1+0.12/12)15 - 1)/(0.12/12) = R 25,755.03

(b) =FV(0.12/12,15,1600,,0) R 25,755.03

Ordinary Annuity Due

(c) Fv = 1600*((1+0.12/12)15-1)*(1+0.12/12)/(0.12/12) = R 26,012.58

(d) =FV(0.12/12,15,1600,,1) R 26,012.58


Exercise 16.23

Compound Interest Fv1 = Pv*(1+0.07)9


Simple Interest Fv2 = Pv*(1+0.07*9)

Difference Fv1 - Fv2 = 334.16

Pv*(1+0.07)9 - Pv*(1+0.07*9) = 334.16


Pv = 334.16/((1+0.07)9 - (1+0.07*9)) = 1602.999

Captial Sum (Pv) = R 1,603.00


Exercise 16.24

(a) Car price in 3 years

Fv (Compound Interest) = 80000*(1+0.04)3 = 89989.12

Invest at end of month


R= 89989.12/(((1+0.09/12)(3*12) - 1)/(0.09/12))
R 2,186.71

(b) Invest at beginning of month


R= 89989.12/(((1+0.09/12)(3*12) - 1)*(1+0.09/12)/(0.09/12))
R 2,170.43
Exercise 16.25

(a) R= 8500/((1-(1+0.18/12)(-3*12))/(0.18/12)) = R 307.30

(b) Total paid = 11062.8

Interest amt = 11062.8 - 8500 = R 2,562.80

% of debt 2562.8/8500% = 30.15%


Exercise 16.26

Present value of an Ordinary Annuity Certain.

Pv = 8750*((1-(1+0.1/12)^(-(5*12)))/(0.1/12)) = R 411,821.98

The employee will receive a gratuity of R411 821.98.


Exercise 16.27

Ordinary Annuity Due

(a) Fv = 750*((1+0.145/4)(4*15)-1)*(1+0.145/4)/(0.145/4) = R 160,149.71

(b) =FV(0.145/4,60,750,,1) (using Excel function) R 160,149.71


Exercise 16.28

Ordinary annuity certain (OAC) for 2 years with R = 540.


(2*12)
Fv1 (OAC) = 540*((1+0.12/12) -1)/(0.12/12) = R14 565.67

Then compute Fv on the capital sum after 2 years until maturity (for 7 years).

Fv1 (CI) = 14565.67*(1+0.12/12)^(7*12) = R33 598.96

Ordinary annuity certain (OAC) for 7 years with R = 750.


(7*12)
Fv2 (OAC) = 750*((1+0.12/12) -1)/(0.12/12) = R98 004.21

Total Funds Available Fv1 (CI) + Fv2 (OAC) = R131 603.17


Exercise 16.29

(a) Ordinary Annuity Certain

Fv (monthly) = 1000*((1+0.085/12)^(1*12)-1)/(0.085/12) = R12 478.72

Fv (quarterly) = 3000*((1+0.10/4)^(1*4)-1)/(0.10/4) = R12 457.55

Conclusion It is better to invest monthly .

(b) Ordinary Annuity Certain - Using Excel 's function, FV.

Fv (monthly) = =FV(0.085/12,12,1000,,0) R12 478.72

Fv (quarterly) = =FV(0.10/4,4,3000,,0) R12 457.55

(c) Ordinary Annuity Due - Using Excel 's function, FV.

Fv (monthly) = =FV(0.085/12,12,1000,,1) R12 567.11

Fv (quarterly) = =FV(0.10/4,4,3000,,1) R12 768.99

Conclusion It is now better to invest quarterly .


Exercise 16.30

(a) PV of an Ordinary Annuity Certain

Pv = 2200*(1-(1+0.09/12)^(-(4*12)))/(0.09/12) = R88 406.52


Deposit = 20000 R20 000,00
Total Purchase Price of Motor Vehicle = R108 406.52

(b) Pv of an Ordinary Annuity Certain - Using Excel 's function, PV

Purchase Price = PV(0.09/12,48,2200,,0) + deposit = R88 406.52 + R20 000


= R108 406.52
Exercise 16.31

(a) Using Ordinary Annuity Certain for 2 years

Fv1 (2 years) = 1000*((1+0.08/12)(2*12)-1)/(0.08/12) = R25 933.19

Using Compound Interest on Capital Sum for 1 year.


FV1 (1 year) =25933.19*(1+0.1/12)(1*12) = R28 648.73

Using Ordinary Annuity Certain for 1 year


(1*12)
Fv2 (1 year) = 1000*((1+0.10/12) -1)/(0.10/12) = R12 565.57
12565.56809

Maturity value after 3 years R28 648.73 + R12 565.57 = R41 214.30

(b) Using Excel 's function FV with a compound interest calculation

=FV(0.08/12,24,1000,,0)*(1+0.1/12)12 + FV(0.1/12,12,1000,,0) R41 214.30


Exercise 16.32

(a) Months 1 - 5
FV1 = 200*((1+0.12/12)5-1)/(0.12/12) = R1020.20
Withdrawal R 300.00
Balance R1020.20 - R300.00 = R 720.20
7
Compound Interest 720.2*(1+0.12/12) = R 772.15

Months 6 - 10
5
FV2 = 200*((1+0.12/12) -1)/(0.12/12) = R1020.20
Withdrawal R 300.00
Balance R1020.20 - R300.00 = R 720.20
Compound Interest 720.2*(1+0.12/12)2 = R 734.68

Months 11, 12
2
FV3 = 200*((1+0.12/12) -1)/(0.12/12) = R 402.00

Total Amount Available after 12 months =


R772.15 + R734.68 + R402.00 = R1908.83

(b) Using Excel 's Function, FV

Months 1 - 5 (Ordinary Annuity Certain - R300 + Compound Interest for 7 months)

7
=(-(FV(0.12/12,5,200,,0))-300)*(1+0.12/12) R 772.15

Months 6 - 10 (Ordinary Annuity Certain - R300 + Compound Interest for 2 months)

2
=(-(FV(0.12/12,5,200,,0))-300)*(1+0.12/12) R 734.68

Months 11, 12 (Ordinary Annuity Certain)

=(-FV(0.01,2,200,,0)) R 402.00

Total Amount Available after 12 months =


R772.15 + R734.68 + R402.00 = R1908.83
Exercise 16.33

(a) Option 1 Repayment at the end of every month


R= 26000/((1-(1+0.14/4)(-3*4))/(0.14/4)) = R2 690.58

(b) Option 2 Repayment at the beginning of every month


R= 26000/(((1-(1+0.14/4)(-3*4))*(1+0.14/4))/(0.14/4)) = R2 599.60

(c) The student should select to repay at the beginning of every month.
Total repayment will be less than repaying at the end of every month.
Exercise 16.34
Present Value (Pv) of a Deferred Annuity

Factor 1 =(1-(1+0.16/12)(-(24+36)))/(0.16/12) 41.12171

(-(36))
Factor 2 =(1-(1+0.16/12) )/(0.16/12) 28.44381

Difference 41.12171 - 28.44381 = 12.6779

R= 30000/12.6779 = R2 366.32

The repayment per month is R2 366.32.


Exercise 16.35

Rate (i ) (half yearly) = (40697.7/18000)(1/10)-1 = 0.085

Rate (i ) (nominal % p.a.) = 0.085*2 = 17% p.a.


Exercise 16.36

Re-arrange the Fv formula for an Ordinary Annuity Certain

Factor 1 (103757.98/2000)*(0.12/12)+1 = 1.51879

n = LOG(1.51879)/LOG(1+0.12/12) = 42 months

The house owner will take 3 years and 6 months to save R103 757.98.
Exercise 16.37

INVESTMENT OPTIONS
Trucking Laundry
Initial investment (R) 60000 60000
Annual Cash Flow (R)
year 1 32000 0
year 2 38500 7500
year 3 26000 45000
year 4 13000 37500
year 5 9500 55500

12% p.a. cost of capital


1 28571 0
2 30692 5979
3 18506 32030
4 8262 23832
5 5391 31492
91422 93333

NPV R31 422 R33 333

Recommend the purchase of the Laundry business.