Professional Documents
Culture Documents
ISTUDY
The McGraw Hill Series in Operations and Decision Sciences
SUPPLY CHAIN MANAGEMENT BUSINESS SYSTEMS DYNAMICS Doane and Seward
Essential Statistics in Business and
Bowersox, Closs, Cooper, and Bowersox Sterman
Economics
Supply Chain Logistics Management Business Dynamics: Systems Thinking
Third Edition
Fifth Edition and Modeling for a Complex World
Lind, Marchal, and Wathen
Johnson OPERATIONS MANAGEMENT
Basic Statistics for Business and
Purchasing and Supply Management
Cachon and Terwiesch Economics
Sixteenth Edition
Operations Management Tenth Edition
Simchi-Levi, Kaminsky, and Simchi-Levi Second Edition
Lind, Marchal, and Wathen
Designing and Managing the Supply
Cachon and Terwiesch Statistical Techniques in Business and
Chain: Concepts, Strategies, Case
Matching Supply with Demand: An Economics
Studies
Introduction to Operations Management Eighteenth Edition
Fourth Edition
Fourth Edition
Jaggia and Kelly
Stock and Manrodt
Jacobs and Chase Business Statistics: Communicating
Fundamentals of Supply Chain
Operations and Supply Chain with Numbers
Management
Management Fourth Edition
PROJECT MANAGEMENT Sixteenth Edition
Jaggia and Kelly
Larson and Gray Jacobs and Chase Essentials of Business Statistics:
Project Management: The Managerial Operations and Supply Chain Communicating with Numbers
Process Management: The Core Second Edition
Eighth Edition Fifth Edition
BUSINESS ANALYTICS
SERVICE OPERATIONS Schroeder and Goldstein
Jaggia, Kelly, Lertwachara, and Chen
MANAGEMENT Operations Management in the Supply
Business Analytics: Communicating
Chain: Decisions and Cases
Bordoloi, Fitzsimmons, and Fitzsimmons with Numbers
Eighth Edition
Service Management: Operations,
BUSINESS MATH
Strategy, Information Technology Stevenson
Ninth Edition Operations Management Slater and Wittry
Fourteenth Edition Practical Business Math Procedures
MANAGEMENT SCIENCE
Thirteenth Edition
Swink, Melnyk, and Hartley
Hillier and Hillier
Managing Operations Across the Supply Slater and Wittry
Introduction to Management Science: A
Chain Math for Business and Finance: An
Modeling and Case Studies Approach
Fourth Edition Algebraic Approach
with Spreadsheets
Second Edition
Sixth Edition BUSINESS STATISTICS
BUSINESS RESEARCH METHODS Bowerman, Drougas, Duckworth, Froelich,
Hummel, Moninger, and Schur
Schindler
Business Statistics and Analytics in
Business Research Methods
Practice
Fourteenth Edition
Ninth Edition
BUSINESS FORECASTING
Doane and Seward
Keating and Wilson Applied Statistics in Business and
Forecasting and Predictive Analytics Economics
Seventh Edition Seventh Edition
ISTUDY
Fourth Edition
BUSINESS STATISTICS
Communicating with Numbers
ISTUDY
Final PDF to printer
BUSINESS STATISTICS
Published by McGraw Hill LLC, 1325 Avenue of the Americas, New York, NY 10121. Copyright © 2022
by McGraw Hill LLC. All rights reserved. Printed in the United States of America. No part of this publication
may be reproduced or distributed in any form or by any means, or stored in a database or retrieval system,
without the prior written consent of McGraw Hill LLC, including, but not limited to, in any network or other
electronic storage or transmission, or broadcast for distance learning.
Some ancillaries, including electronic and print components, may not be available to customers outside the
United States.
1 2 3 4 5 6 7 8 9 LWI 24 23 22 21
ISBN 978-1-260-59756-1
MHID 1-260-59756-3
All credits appearing on page or at the end of the book are considered to be an extension of the copyright page.
The Internet addresses listed in the text were accurate at the time of publication. The inclusion of a website
does not indicate an endorsement by the authors or McGraw Hill LLC, and McGraw Hill LLC does not
guarantee the accuracy of the information presented at these sites.
mheducation.com/highered
ISTUDY
A B O U T T H E AU T H O R S
Sanjiv Jaggia
Sanjiv Jaggia is a professor of economics and finance at
California Polytechnic State University in San Luis Obispo.
Dr. Jaggia holds a Ph.D. from Indiana University and is a
Chartered Financial Analyst (CFA®). He enjoys research
in statistics and data analytics applied to a wide range of
business disciplines. Dr. Jaggia has published numerous
papers in leading academic journals and has co-authored
three successful textbooks, two in business statistics and
Courtesy Sanjiv Jaggia
one in business analytics. His ability to communicate in the
classroom has been acknowledged by several teaching
awards. Dr. Jaggia resides in San Luis Obispo with his wife and daughter. In his spare
time, he enjoys cooking, hiking, and listening to a wide range of music.
Alison Kelly
Alison Kelly is a professor of economics at Suffolk University in Boston. Dr. Kelly holds a Ph.D.
from Boston College and is a Chartered Financial Analyst (CFA®). Dr. Kelly has published in
a wide variety of academic journals and has co-authored three successful textbooks, two
in business statistics and one in business analytics. Her
courses in applied statistics and econometrics are popular
with students as well as working professionals. She has
also served as a consultant for a number of companies;
her most recent work focused on how large financial
institutions satisfy requirements mandated by the Dodd-
Frank Act. Dr. Kelly resides in Hamilton, Massachusetts,
with her husband, daughter, and son. In her spare time,
Courtesy Alison Kelly
she enjoys exercising and gardening.
vi
ISTUDY
Business Statistics: Communicating
with Numbers
Reviewer Quotes
WALKTHROUGH
ISTUDY
A Unique Emphasis on
Communicating with Numbers
Makes Business Statistics Relevant
to Students
We wrote Business Statistics: Communicating with Numbers because we saw a need
for a contemporary, core statistics text that sparked student interest and bridged the gap
between how statistics is taught and how practitioners think about and apply statistical
methods. Throughout the text, the emphasis is on communicating with numbers rather
than on number crunching. In every chapter, students are exposed to statistical informa-
tion conveyed in written form. By incorporating the perspective of practitioners, it has
been our goal to make the subject matter more relevant and the presentation of material
more straightforward for students. Although the text is application-oriented and practical,
it is also mathematically sound and uses notation that is generally accepted for the topic
being covered.
From our years of experience in the classroom, we have found that an effective way
to make statistics interesting is to use timely applications. For these reasons, examples
in Business Statistics come from all walks of life, including business, economics, sports,
health, housing, the environment, polling, and psychology. By carefully matching exam-
ples with statistical methods, students learn to appreciate the relevance of statistics in our
world today, and perhaps, end up learning statistics without realizing they are doing so.
Integrated Introductory Cases. Each chapter begins with an interesting and relevant
introductory case. The case is threaded throughout the chapter, and once the relevant sta-
tistical tools have been covered, a synopsis—a short summary of findings—is provided.
The introductory case often serves as the basis of several examples in other chapters.
Writing with Data. Interpreting results and conveying information effectively is critical
to effective decision making in a business environment. Students are taught how to take
the data, apply it, and convey the information in a meaningful way.
Written as Taught. Topics are presented the way they are taught in class, beginning
with the intuition and explanation and concluding with the application.
Connect. Connect is an online system that gives students the tools they need to be suc-
cessful in the course. Through guided examples and LearnSmart adaptive study tools,
students receive guidance and practice to help them master the topics.
ISTUDY
Features New to the Fourth Edition
In the fourth edition of Business Statistics, we have made substantial revisions that
address the current needs of the market. These revisions are based on the feedback of
countless reviewers and users of our earlier editions.
The emphasis in this edition has been to strengthen the connection between business
statistics and data analytics. More than ever, colleges and universities across the United
States and abroad are incorporating business analytics into their curricula, and businesses
are scrambling to find qualified professionals who can translate statistical analysis into
decisions that improve performance. We believe that the fourth edition will not only
introduce students to data analytics, but will also excite them to further explore the field.
There are four major innovations in this edition.
WALKTHROUGH
ISTUDY
Students Learn Through Real-World
Cases and Business Examples . . . Rev. Confirming Pages
Integrated Introductory Cases The numerator of the Sharpe ratio measures the extra reward that investors receive for the
added risk taken—this difference is often called excess return. The higher the Sharpe
ratio, the better the investment compensates its investors for risk.
Each chapter opens with a real-life case study that forms the basis for several examples
within the chapter. The questions included in the
EXAMPLE examples
3.19 create a roadmap for master-
ing the most important learning outcomes within the chapter. A onsynopsis
Growth and Value given that the return of each chap-
Use the information in Table 3.13 to calculate and interpret the Sharpe ratios for
a 1-year T-bill is 2%.
Investment Decision
continue, the investor will favor investing in Growth over Value. A commonly used disclaimer, however, states that past
performance is no guarantee of future results.
Dorothy Brennan works as a financial advisor at a large investment firm. She meets with an inex-
perienced investor who has some questions regarding two approaches to mutual fund investing:
growth investing versus value investing. The investor has heard that growth funds invest in com-
108 B u S I n e S S S TAT I S T I C S 3.5 Mean-Variance Analysis and the Sharpe Ratio
panies whose stock prices are expected to grow at a faster rate, relative to the overall stock mar-
ket. Value funds, on the other hand, invest in companies whose stock prices are below their true
worth. The investor has also heard that the main component of investment return is through capi-
tal appreciation in growth funds and through dividend income in value funds.
The investor shows Dorothy the annual return data for Fidelity’s Growth Index mutual fund
(Growth) and Fidelity’s Value Index mutual fund (Value). Table 3.1 shows a portion jag16309_ch03_080-123
of the annual108 07/22/20 04:00 PM
return (in %) for these two mutual funds from 1984 to 2019. It is difficult for the investor to draw
any conclusions from the data in its present form. In addition to clarifying the style differences
in growth investing versus value investing, the investor requests Dorothy to summarize
Growth_Value
the data.
TABLE 3.1 Annual Returns (in %) for Growth and Value
Year Growth Value
1984 −5.50 −8.59
1985 39.91 22.10
⋮ ⋮ ⋮
2019 38.42 31.62
81
x B U S I N E S S S TAT I S T I C S WALKTHROUGH
ISTUDY
and Build Skills to Communicate
Results
Writing with Data
One of our most important innovations is the inclusion of a sample report within every
chapter. Our intent is to show students how to convey statistical information in written
form to those who may not know detailed statistical methods. For example, such a report
may be needed as input for managerial decision making in sales, marketing, or company
planning. Several similar writing exercises are provided at the end of every Writing with
Data section. Each chapter also includes a synopsis that addresses questions raised from
the introductory case. This serves as a shorter writing sample for students. Instructors
Confirming Pages
of large sections may find these reports useful for incorporating writing into their
statistics courses.
Confirming Pages
6 . 4 W R I T I N G W I T H DATA
Case Study
Professor Lang is a professor of economics at Salem State university. She has been
6 . 4teaching
W R IaTcourse
I N Gin W
Principles
I T H ofD economics
ATA for over 25 years. Professor Lang has
never graded on a curve since she believes that relative grading may unduly penalize
Case Study
(benefit) a good (poor) student in an unusually strong (weak) class. She always uses an
absolute scale for making grades, as shown in the two left columns of table 6.4.
Professor Lang is a professor of economics at Salem State University. She has been
teaching a course in Principles
TABLE of Economics
6.4 Grading for over
Scales with 25 years.
Absolute GradingProfessor Lang has
versus Relative Grading
never graded on a curve since she believes that relative grading may unduly penalize
Absolute
(benefit) a good (poor) student Grading
in an unusually strong (weak) class. SheRelative Grading
always uses an
absolute scale forGrade
making grades, as shown
Scorein the two left columns
Gradeof Table 6.4. Probability
A Grading Scales with Absolute
TABLE 6.4 92 and above
Grading A Grading
versus Relative 0.10
B
Absolute Grading 78 up to 92 Relative
B Grading 0.35
Grade C Score 64 up Grade
to 78 C Probability 0.40
A D 92 and above
58 up Ato 64 D 0.10 0.10 Jasper White/image Source
B 78 up to 92 B 0.35
F Below 58 F 0.05
C 64 up to 78 C 0.40
D A colleague of58 up to 64 Lang’s D has convinced her 0.10 Image Source, all rights reserved.
Professor to move to relative grading, because it
F
corrects Below 58 problems.
for unanticipated F Professor Lang decides 0.05 to experiment with grading based
on the relative
A colleague scaleLang’s
of Professor as shown in the two
has convinced herright columns
to move of table
to relative 6.4.
grading, usingitthis relative grading
because
scheme,
corrects the top 10%
for unanticipated of students
problems. Professorwill
Langget A’s, the
decides next 35%with
to experiment B’s,grading
and sobased on. Based on her years
on theofrelative scale experience,
teaching as shown in the two right columns
Professor of Table 6.4.
Lang believes thatUsing this relative
the scores grading
in her course follow a normal
scheme, the top 10%with
distribution of students
a meanwillof get
78.6A’s,and
the a
next 35% B’s,deviation
standard and so on.of Based
12.4.on her years
of teachingProfessor
experience, Professor Lang believes that the scores in her course follow a normal
Lang wants to use this information to calculate probabilities based on the absolute
distribution with a mean of 78.6 and a standard deviation of 12.4.
scale and compare them to the probabilities based on the relative scale. then, she wants to cal-
Professor Lang wants to use this information to calculate probabilities based on the absolute
culate the range
scale and compare them to ofthe
scores for grades
probabilities basedbased
on the on the relative
relative scale
scale. Then, sheand
wantscompare
to cal- them to the abso-
culate lute scale.ofFinally,
the range she
scores for wantbased
grades to determine whichscale
on the relative grading scale makes
and compare them toit the
harder
abso-to get higher grades.
lute scale. Finally, she want to determine which grading scale makes it harder to get higher grades.
Sample
Many teachers would confess that grading is one of the most difficult tasks of their profession.
Sample
Report—
Many teachers would confess that grading is one of the most difficult tasks of their profession.
two common
Two common gradingused
grading systems systems used
in higher in higher
education are education
relative and are relative
absolute. Report—
and absolute. Relative
Relative
grading
systemssystems are norm-referenced or curve-based, in iswhich
basedaongrade
the is based on the
grading are norm-referenced
position in
or curve-based,
class. Absolute
in which a grade
grading systems, on the other Absolute
student’s relative position in class. Absolute grading systems, on the other hand, are criterion- hand, are criterion-
student’s relative Absolute
referenced,
short, short,
in whichin
referenced,
with absolute
a grade
whichisarelated
grading, the
with absolute
gradetoisthe
student’s
grading,
student’s
related
thescore
absolute
to the
is compared
student’s
performance
student’s absolute
scoretoisa compared
predetermined
in class. In
performance
to ascale,
Grading
in class. in
predetermined scale,
Grading
whereas with relative grading, the score is compared to the scores of other students in
whereas with relative grading, the score is compared to the scores of other students in versus versus
the class.
Letthe class. a grade in Professor Lang’s class, which is normally distributed with a mean Relative
X represent
Relative
of 78.6 andLet X represent
a standard a grade
deviation inThis
of 12.4. Professor Lang’s
information class,
is used whichthe
to derive is grade
normally
Grading
distributed with a mean
probabili-
ties based
P(X ≥ 92) =
on the
of 78.6
P(Z
and
≥
absolute
1.08) =
scale.deviation
a standard
0.14.
For instance,
Other
the probability
of 12.4.
probabilities, derived
of receiving
this information
similarly, are
is an
used
presented
ties based on the absolute scale. For instance, the probability of receiving
A isto
in
derived
deriveasthe grade probabili-
Table 6.5. an A is derived as
Grading
P(X ≥ 92) = P(Z ≥ 1.08) = 0.14. Other probabilities, derived similarly, are presented in table 6.5.
TABLE 6.5 Probabilities Based on Absolute Scale and Relative Scale
Probability
TABLE Based on Based
6.5 Probabilities Probability Based
on Absolute onand Relative Scale
Scale
Grade Absolute Scale Relative Scale
A Probability Based on
0.14 0.10 Probability Based on
B Grade 0.38 Absolute Scale 0.35 Relative Scale
C A 0.36 0.14 0.40 0.10
D B 0.07 0.38 0.10 0.35
F C 0.05 0.36 0.05 0.40
D 0.07 0.10
F 0.05 0.05
229
229
WALKTHROUGH
ISTUDY
Unique Coverage and
Presentation . . .
Unique Coverage of Regression Analysis
Our coverage of regression analysis is more extensive than that of the vast majority of texts.
This focus reflects the topic’s importance in the emerging field of data analytics. We combine
simple and multiple regression in one chapter, which we believe is a seamless grouping and
eliminates needless repetition. The detailed Excel and R instructions eliminate the need for
tedious manual calculations. Three more in-depth chapters cover statistical inference, nonlinear
relationships, dummy variables, and the linear probability and the logistic regression models.
The emphasis in all chapters is on conceptualization, model selection, and interpretation of
the results with reference to professionally created figures and tables.
Chapter 14: Regression Analysis
Chapter 15: Inference with Regression Models
Chapter 16: Regression Models for Nonlinear Relationships
Chapter 17: Regression Models with Dummy Variables
The authors have put forth a novel and innovative way to present
regression which in and of itself should make instructors take a long
and hard look at this book. Students should find this book very
readable and a good companion for their course.
Harvey A. Singer, George Mason University
Written as Taught
We introduce topics just the way we teach them; that is, the relevant tools follow the
opening application. Our roadmap for solving problems is
1. Start with intuition
2. Use Excel or R to estimate the appropriate model,
3. Communicate the results.
We use worked examples throughout the text to illustrate how to apply concepts to solve
real-world problems.
ISTUDY
that Make the Content More
Effective
EXAMPLE 5.9
People turn to social media to stay in touch with friends and family members, catch
the news, look for employment, and be entertained. According to a recent survey,
Integration of Microsoft
68% of all U.S. Excel
adults are Facebook and
users. R a sample of 100 randomly
Consider
selected
We prefer thatAmerican adults.
students first focus on and absorb the statistical material before replicating
their a.
results
Whatwith a computer.
is the probabilitySolving each70application
that exactly manually
American adults provides students with
are Facebook
a deeperusers?
understanding of the relevant concept. However, we recognize that embedding
computer
b. Whatoutput is probability
is the often necessary
that noinmore
order to 70
than avoid cumbersome
American calculations or the
adults are
need for Facebook
statistical users?
tables. Microsoft Excel and R are the primary software packages used
in this
c. text.
WhatWe is chose Excel and
the probability thatRatover
leastother statisticaladults
70 American packages based onusers?
are Facebook their wide-
spread use and reviewer feedback. For instructors who prefer to focus only on Excel, the
R instructions sections are easily skipped. We provide brief guidelines for using Minitab,
SPSS,SOLUTION:
and JMPWe in let
chapter appendices.
X denote the number of American adults who are Facebook users.
We also know that p = 0.68 and n = 100.
Using Excel
We use Excel’s BINOM.DIST function to calculate binomial probabilities. We
enter =BINOM.DIST(x, n, p, TRUE or FALSE) where x is the number of suc-
cesses, n is the number of trials, and p is the probability of success. For the last
argument, we enter TRUE if we want to find the cumulative probability function
P(X ≤ x) or FALSE if we want to find the probability mass function P(X = x).
a. In order to find the probability that exactly 70 American adults are Facebook
users, P(X = 70), we enter =BINOM.DIST(70, 100, 0.68, FALSE) and Excel
returns 0.0791.
b. In order to find the probability that no more than 70 American adults are
Facebook users, P(X ≤ 70), we enter =BINOM.DIST(70, 100, 0.68, TRUE)
and Excel returns 0.7007.
c. In order to find the probability that at least 70 American adults are Facebook
users, P(X ≥ 70) = 1 − P(X ≤ 69), we enter =1–BINOM.DIST(69, 100, 0.68,
TRUE) and Excel returns 0.3784.
Using R
We use R’s dbinom and pbinom functions to calculate binomial probabilities. In
order to calculate the probability mass function P(X = x), we enter dbinom(x, n, p)
where x is the number of successes, n is the number of trials, and p is the probabil-
ity of success. In order to calculate the cumulative probability function P(X ≤ x),
we enter pbinom(x, n, p).
a. In order to find P(X = 70), we enter:
> dbinom(70, 100, 0.68)
And R returns: 0.07907911.
b. In order to find P(X ≤ 70), we enter:
> pbinom(70, 100, 0.68)
And R returns: 0.7006736.
c. In order to find P(X ≥ 70) = 1 − P(X ≤ 69), we enter:
> 1 − pbinom(69, 100, 0.68)
And R returns: 0.3784055.
ISTUDY
Real-World Exercises and Case
Studies that Reinforce the Material
Mechanical and Applied Exercises
Chapter exercises are a well-balanced blend of mechanical, computational-type problems
followed by more ambitious, interpretive-type problems. We have found that simpler
drill problems tend to build students’ confidence prior to tackling more difficult applied
problems. Moreover, we repeatedly use many data sets—including house prices, sales,
personality types, health measures, expenditures, stock returns, salaries, and debt—in
various chapters of the text. For instance, students first use these real Rev.
dataConfirming Pages
to calculate
summary measures, make statistical inferences with confidence intervals and hypothesis
tests, and finally, perform regression analysis.
E X E RC I S E S 3.4
Mechanics 44. StockPrices. Monthly closing stock prices for Firm A
39. Consider the following population data: and Firm B are collected for the past five years. A portion of
the data is shown in the accompanying table.
34 42 12 10 22
Observation Firm A Firm B
a. Calculate the range. 1 39.91 42.04
b. Calculate MAD.
2 42.63 41.64
c. Calculate the population variance.
⋮ ⋮ ⋮
d. Calculate the population standard deviation.
60 87.51 75.09
40. Consider the following population data:
a. Calculate the sample variance and the sample standard
0 −4 2 −8 10 deviation for each firm’s stock price.
a. Calculate the range. b. Which firm’s stock price had greater variability as
measured by the standard deviation?
b. Calculate MAD.
c. Which firm’s stock price had the greater relative
c. Calculate the population variance.
dispersion?
d. Calculate the population standard deviation.
45. Rental. A real estate analyst examines the rental mar-
41. Consider the following sample data:
ket in a college town. She gathers data on monthly rent and
40 48 32 52 38 42 the square footage for 40 apartments. A portion of the data is
shown in the accompanying table.
a. Calculate the range.
b. Calculate MAD. Monthly Rent Square Footage
c. Calculate the sample variance. 645 500
d. Calculate the sample standard deviation. 675 648
42. Consider the following sample data: ⋮ ⋮
2,400 2,700
− 10 12 −8 −2 −6 8
a. Calculate the mean and the standard deviation for
a. Calculate the range. monthly rent.
b. Calculate MAD. b. Calculate the mean and the standard deviation for square
c. Calculate the sample variance and the sample standard footage.
deviation. c. Which sample data exhibit greater relative dispersion?
46. Revenues. The accompanying data file shows the
Applications annual revenues (in $ millions) for Corporation A and Corpora-
43. Prime. The accompanying table shows a portion of the tion B for the past 13 years.
annual expenditures (in $) for 100 Prime customers. a. Calculate the coefficient of variation for Corporation A.
b. Calculate the coefficient of variation for Corporation B.
Customer Expenditures c. Which variable exhibits greater relative dispersion?
1 1272 47. Census. The accompanying data file shows, among
2 1089 other variables, median household income and median house
⋮ ⋮ value for the 50 states.
100 1389 a. Calculate and discuss the range of household income
and house value.
a. What were minimum expenditures? What were maximum b. Calculate the sample MAD and the sample standard
expenditures? deviation of household income and house value.
b. Calculate the mean and the median expenditures. c. Discuss why we cannot directly compare the sample
c. Calculte the variance and the standard devation. MAD and the standard deviations of the two variables.
ISTUDY
Suggested Case Studies
Here are some suggestions for analysis.
Typical Report 14.2 House_Price. Choose a college town. Find the model that best predicts
the sale price of a house. Use goodness-of-fit measures to find the appropriate explanatory
variables.
First Pages
Conceptual
Report 14.3
Review
enrolled students
College_Admissions. Choose a college of interest and use the sample of
to best predict a student’s college grade point average. Use goodness-of-fit
measures to find the appropriate explanatory variables. In order to estimate these models, you
At the endhave
oftoeach chapter, we present a conceptual review that provides a more holistic
first filter the data to include only the enrolled students.
approach Report
to reviewing
14.4
the material. This section revisits the learning outcomes and pro-
NBA. Develop and compare two models for predicting a player’s salary
vides the most
modelonimportant
based uses
offensive onedefinitions,
only performance
explanatory
measuresinterpretations,
variable to predict
and defensive anddescribe
and/or
performance formulas.
changes
measures, in the
respectively.
response
In order to variable. The models,
estimate these model isyouexpressed as filter
have to first β0 +data
y = the β1xto+include
ε, where and xstatistics
onlyycareer are the
response
based variable
on regular and theexclude
seasons. explanatory
playersvariable, respectively,
with no information and ε is the random error
on salary.
term. The coefficients β0 and β1 are the unknown parameters to be estimated.
We apply the ordinary least squares (OLS) method to find a sample regression equation
ŷ = b0 + b1x, where ŷ is the predicted value of the response variable and b0 and b1 are the
estimates of β0 and β1, respectively. The estimated slope coefficient b1 represents the
C O nin C
change e Px Tchanges
ŷ when U A byL oneRunit.
e VTheI eunits
Wof b1 are the same as those of y.
LO Estimate aand
14.3 Conduct
LO 14.1 interprettest
hypothesis the for
multiple linear regression
the population model.
correlation coefficient.
The multiple
When linear
conducting regression
tests concerningmodel allows
the population
____
more than
correlation one explanatory
coefficient ρ variable
xy to of
, the value be
linearly related with the response variable rxy√n y. 2 is defined as y = β0 + β1x1 + β2x2 + ⋯ +
− It
the _________
βkxktest
+ ε,statistic
where isy is
calculated as tdf variable,
the response = _____ x1, x, 2where is the
, . . ., xrkxyare the sample correlation
k explanatory coef-
variables,
and ε isand
ficient thedfrandom √1 − r 2xy β0, β1, . . ., βk are the unknown parame-
error term. The coefficients
= n − 2.
ters to be estimated. We again use the OLS method to arrive at the following sample
regression equation: ŷand
LO 14.2 Estimate 0 + b1x1 +the
= binterpret + ⋯ +linear
b2x2simple bkxk, where b0, b1, model.
regression . . ., bk are the esti-
mates of β0, β1, . . ., βk, respectively.
Regression analysis presumes that one variable, called the response variable, is influ-
For each
enced by explanatory variable
other variables, (j =explanatory
calledxj the 1, . . ., k), the corresponding
variables. The simpleslopelinear
coefficient bj is
regression
the estimated regression coefficient. It measures the change in the predicted value of the
response variable ŷ, given a unit increase in the associated explanatory variable xj, hold-
ing all other explanatory variables constant. In other words, it represents the partial
influence of xj on ŷ. CHAPTeR 14 Regression Analysis BUSIneSS S TAT I S T I C S 513
ADDITIOnAL eXeRCISeS
59. Energy_Health. The following table shows a b. Specify the competing hypotheses in order to
portion of the monthly returns data (in percent) for determine whether the population correlation
2010–2016 for two of Vanguard’s mutual funds: the coefficient is different from zero.
Energy Fund and the Healthcare Fund. c. At the 5% significance level, what is the
conclusion to the test? Are the returns on the
Date Energy Healthcare mutual funds correlated?
Jan-10 −4.86 −0.13 60. Yields. While the Federal Reserve controls
Feb-10 1.50 0.58 short-term interest rates, long-term interest
⋮ ⋮ ⋮ rates essentially depend on supply and demand
Dec-16 −0.30 −5.26 dynamics, as well as longer-term interest rate
expectations. The accompanying table shows
a. Calculate and interpret the sample correlation a portion of data for the annualized rates for
coefficient rxy. the 10-year Treasury yield (in %) and the
WALKTHROUGH
ISTUDY
Instructors: Student Success Starts with You
Tools to enhance your unique voice
Want to build your own course? No problem. Prefer to use our
turnkey, prebuilt course? Easy. Want to make changes throughout the
65%
Less Time
semester? Sure. And you’ll save time with Connect’s auto-grading too.
Grading
ISTUDY
Students: Get Learning that Fits You
Effective tools for efficient studying
Connect is designed to make you more productive with simple, flexible, intuitive tools that maximize
your study time and meet your individual learning needs. Get learning that works for you with Connect.
ISTUDY
Remote Proctoring & Browser-Locking Capabilities
Seamlessly integrated within Connect, these services allow instructors to control stu-
dents’ assessment experience by restricting browser activity, recording students’ activity,
and verifying students are doing their own work.
Instant and detailed reporting gives instructors an at-a-glance view of potential academic
integrity concerns, thereby avoiding personal bias and supporting evidence-based claims.
Tegrity Campus:
Lectures 24/7
Tegrity Campus is integrated in Connect to help make your class time available 24/7.
With Tegrity, you can capture each one of your lectures in a searchable format for stu-
dents to review when they study and complete assignments using Connect. With a simple
one-click start-and-stop process, you can capture everything that is presented to students
during your lecture from your computer, including audio. Students can replay any part of
any class with easy-to-use browser-based viewing on a PC or Mac.
Educators know that the more students can see, hear, and experience class resources,
the better they learn. In fact, studies prove it. With Tegrity Campus, students quickly
recall key moments by using Tegrity Campus’s unique search feature. This search helps
students efficiently find what they need, when they need it, across an entire semester of
class recordings. Help turn all your students’ study time into learning moments immediately
supported by your lecture. To learn more about Tegrity, visit http://tegritycampus.mhhe.com.
ISTUDY
ALEKS
ALEKS is an assessment and learning program that provides individualized instruction
in Business Statistics, Business Math, and Accounting. Available online in partnership
with McGraw Hill, ALEKS interacts with students much like a skilled human tutor, with
the ability to assess precisely a student’s knowledge and provide instruction on the exact
topics the student is most ready to learn. By providing topics to meet individual stu-
dents’ needs, allowing students to move between explanation and practice, correcting and
analyzing errors, and defining terms, ALEKS helps students to master course content
quickly and easily.
ALEKS also includes an instructor module with powerful, assignment-driven fea-
tures and extensive content flexibility. ALEKS simplifies course management and allows
instructors to spend less time with administrative tasks and more time directing student
learning. To learn more about ALEKS, visit www.aleks.com.
WALKTHROUGH
ISTUDY
What Resources are Available for
Students?
Rev. Confirming Pages
a.
34 42
in the chapter. The video shows the steps to take to work through an exercise. Students can go through each example
multiple times if needed.
The Connect Student Resource page is the place for students to access additional resources. The Student Resource
page offers students quick access to the recommended study tools, data files, and helpful tutorials on statistical programs.
xx B U S I N E S S S TAT I S T I C S WALKTHROUGH
ISTUDY
McGraw Hill Customer Care
Contact Information
At McGraw Hill, we understand that getting the most from new technology can be chal-
lenging. That’s why our services don’t stop after you purchase our products. You can
e-mail our product specialists 24 hours a day to get product training online. Or you can
search our knowledge bank of frequently asked questions on our support website.
For customer support, call 800-331-5094 or visit www.mhhe.com/support. One of our
technical support analysts will be able to assist you in a timely fashion.
WALKTHROUGH
ISTUDY
AC K N O W L E D G M E N T S
We would like to acknowledge the following people for providing useful comments and
suggestions for past and present editions of all aspects of Business Statistics.
John Affisco Hofstra University Sangit Chatterjee Northeastern Soheila Fardanesh Towson
Mehdi Afiat College of University University
Southern Nevada Leida Chen California Carol Flannery University of
Mohammad Ahmadi University Polytechnic State University Texas–Dallas
of Tennessee–Chattanooga Anna Chernobai Syracuse Sydney Fletcher Mississippi
Sung Ahn Washington State University Gulf Coast Community
University Alan Chesen Wright State College
Mohammad Ahsanullah Rider University Andrew Flight Portland State
University Juyan Cho Colorado State University
Imam Alam University of University–Pueblo Samuel Frame Cal Poly San
Northern Iowa Alan Chow University of South Luis Obispo
Mostafa Aminzadeh Towson Alabama Priya Francisco Purdue
University Bruce Christensen Weber State University
Ardavan Asef-Vaziri California University Vickie Fry Westmoreland
State University Howard Clayton Auburn County Community College
Antenah Ayanso Brock University Ed Gallo Sinclair Community
University Robert Collins Marquette College
Scott Bailey Troy University University Glenn Gilbreath Virginia
Jayanta Bandyopadhyay Central M. Halim Dalgin Kutztown Commonwealth University
Michigan University University Robert Gillette University of
Samir Barman University of Tom Davis University of Kentucky
Oklahoma Dayton Xiaoning Gilliam Texas Tech
Douglas Barrett University of Matthew Dean University of University
North Alabama Maine Mark Gius Quinnipiac
John Beyers University of Jason Delaney University of University
Maryland Arkansas–Little Rock Malcolm Gold Saint Mary’s
Arnab Bisi Purdue University– Ferdinand DiFurio Tennessee University of Minnesota
West Lafayette Tech University Michael Gordinier Washington
Gary Black University of Matt Dobra UMUC University
Southern Indiana Luca Donno University of Deborah Gougeon University of
Randy Boan Aims Community Miami Scranton
College Joan Donohue University of Don Gren Salt Lake Community
Matthew Bognar University of South Carolina College
Iowa David Doorn University of Thomas G. Groleau Carthage
Juan Cabrera Ramapo College Minnesota College
of New Jersey James Dunne University of Babita Gupta CSU Monterey
Scott Callan Bentley University Dayton Bay
Gregory Cameron Brigham Mike Easley University of New Robert Hammond North
Young University Orleans Carolina State University
Kathleen Campbell St. Joseph’s Erick Elder University of Sheila Diann Hammon Athens
University Arkansas–Little Rock State University
Alan Cannon University of Ashraf ElHoubi Lamar Jim Han Florida Atlantic
Texas–Arlington University University
Michael Cervetti University of Roman Erenshteyn Goldey- Elizabeth Haran Salem State
Memphis Beacom College University
Samathy Chandrashekar Kathryn Ernstberger Indiana Jack Harshbarger Montreat
Salisbury University University College
Gary Huaite Chao University of Grace Esimai University of Edward Hartono University of
Pennsylvania–Kutztown Texas–Arlington Alabama–Huntsville
xxii
ISTUDY
Clifford Hawley West Virginia Carin Lightner North Carolina Robert Nauss University of
University A&T State University Missouri–St. Louis
Santhi Heejebu Cornell College Constance Lightner Fayetteville Satish Nayak University of
Paul Hong University of Toledo State University Missouri–St. Louis
Ping-Hung Hsieh Oregon State Scott Lindsey Dixie State Thang Nguyen California State
University College of Utah University–Long Beach
Marc Isaacson Augsburg College Ken Linna Auburn Mohammad Oskoorouchi
Mohammad Jamal Northern University–Montgomery California State University–
Virginia Community College Andy Litteral University of San Marcos
Robin James Harper College Richmond Barb Osyk University of Akron
Molly Jensen University of Jun Liu Georgia Southern Bhavik Pathak Indiana
Arkansas University University South Bend
Craig Johnson Brigham Young Chung-Ping Loh University of Melissa Patterson Chabot
University–Idaho North Florida College
Janine Sanders Jones University Salvador Lopez University of Scott Paulsen Illinois Central
of St. Thomas West Georgia College
Vivian Jones Bethune– John Loucks St. Edward’s James Payne Calhoun
Cookman University University Community College
Yogesh Joshi Kingsborough Cecilia Maldonado Georgia Norman Pence Metropolitan
Community College Southwestern State State College of Denver
Jerzy Kamburowski University University Dane Peterson Missouri State
of Toledo Farooq Malik University of University
Howard Kaplon Towson Southern Mississippi Joseph Petry University of
University Brent Marinan University of Illinois–Urbana/Champaign
Krishna Kasibhatla North Arizona Courtney Pham Missouri State
Carolina A&T State Ken Mayer University of University
University Nebraska–Omaha Martha Pilcher University of
Mohammad Kazemi University Bradley McDonald Northern Washington
of North Carolina–Charlotte Illinois University Cathy Poliak University of
Ken Kelley University of Notre Elaine McGivern Duquesne Wisconsin–Milwaukee
Dame University Simcha Pollack St. John’s
Lara Khansa Virginia Tech John McKenzie Babson University
Esther C. Klein St. Francis University Hamid Pourmohammadi
College Norbert Michel Nicholls State California State University–
Ronald Klimberg St. Joseph’s University Dominguez Hills
University John Miller Sam Houston State Tammy Prater Alabama State
Andrew Koch James Madison University University
University Virginia Miori St. Joseph’s Zbigniew H. Przasnyski Loyola
Subhash Kochar Portland State University Marymount University
University Prakash Mirchandani University Manying Qiu Virginia State
Brandon Koford Weber of Pittsburgh University
University Jason Molitierno Sacred Heart Troy Quast Sam Houston State
Randy Kolb St. Cloud State University University
University Elizabeth Moliski University of Michael Racer University of
Vadim Kutsyy San Jose State Texas–Austin Memphis
University Joseph Mollick Texas A&M Srikant Raghavan Lawrence
Francis Laatsch University of University–Corpus Christi Technological University
Southern Mississippi James Moran Oregon State Bharatendra Rai University of
David Larson University of University Massachusetts–Dartmouth
South Alabama Khosrow Moshirvaziri Michael Aaron Ratajczyk
John Lawrence California State California State University– Saint Mary’s University of
University–Fullerton Long Beach Minnesota
Shari Lawrence Nicholls State Tariq Mughal University of Tony Ratcliffe James Madison
University Utah University
Radu Lazar University of Patricia Mullins University of David Ravetch University of
Maryland Wisconsin–Madison California
David Leupp University of Kusum Mundra Rutgers Bruce Reinig San Diego State
Colorado–Colorado Springs University–Newark University
Carel Ligeon Auburn Anthony Narsing Macon State Darlene Riedemann Eastern
University–Montgomery College Illinois University
ACKNOWLEDGEMENTS
ISTUDY
David Roach Arkansas Tech Gary Smith Florida State Ahmad Vakil Tobin College of
University University Business
Carolyn Rochelle East Antoinette Somers Wayne State Tim Vaughan University of
Tennessee State University University Wisconsin–Eau Claire
Alfredo Romero North Ryan Songstad Augustana Raja Velu Syracuse University
Carolina A&T State College Holly Verhasselt University of
University Erland Sorensen Bentley Houston–Victoria
Ann Rothermel University of University Zhaowei Wang Citizens Bank
Akron Arun Kumar Srinivasan Rachel Webb Portland State
Jeff Rummel Emory University Indiana University–Southeast University
Deborah Rumsey The Ohio Anne-Louise Statt University of Kyle Wells Dixie State College
State University Michigan–Dearborn Alan Wheeler University of
Stephen Russell Weber State Scott Stevens James Madison Missouri–St. Louis
University University Mary Whiteside University of
William Rybolt Babson College Alicia Strandberg Temple Texas–Arlington
Fati Salimian Salisbury University University Blake Whitten University of Iowa
Fatollah Salimian Perdue Linda Sturges Suny Maritime Rick Wing San Francisco State
School of Business College University
Samuel Sarri College of Wendi Sun Rockland Trust Jan Wolcott Wichita State
Southern Nevada Bedassa Tadesse University of University
Jim Schmidt University of Minnesota Rongning Wu Baruch College
Nebraska–Lincoln Pandu Tadikamalta University John Yarber Northeast
Patrick Scholten Bentley of Pittsburgh Mississippi Community
University Roberto Duncan College
Bonnie Schroeder Ohio State Tarabay University of John C. Yi St. Joseph’s
University Wisconsin–Madison University
Sue Schou Boise State Faye Teer James Madison Kanghyun Yoon University of
University University Central Oklahoma
Pali Sen University of North Deborah Tesch Xavier Mark Zaporowski Canisius
Florida University College
Donald Sexton Columbia Patrick Thompson University of Ali Zargar San Jose State
University Florida University
Vijay Shah West Virginia Satish Thosar University of Dewit Zerom California State
University–Parkersburg Redlands University
Dmitriy Shaltayev Christopher Ricardo Tovar-Silos Lamar Eugene Zhang Midwestern
Newport University University State University
Soheil Sibdari University of Quoc Hung Tran Bridgewater Ye Zhang Indiana
Massachusetts–Dartmouth State University University–Purdue
Prodosh Simlai University of Elzbieta Trybus California University–Indianapolis
North Dakota State University–Northridge Yi Zhang California State
Harvey Singer George Mason Fan Tseng University of University–Fullerton
University Alabama–Huntsville Yulin Zhang San Jose State
Harry Sink North Carolina Silvanus Udoka North Carolina University
A&T State University A&T State University Qiang Zhen University of North
Don Skousen Salt Lake Shawn Ulrick Georgetown Florida
Community College University Wencang Zhou Baruch College
Robert Smidt California Bulent Uyar University of Zhen Zhu University of Central
Polytechnic State University Northern Iowa Oklahoma
The editorial staff of McGraw Hill are deserving of our gratitude for their guidance
throughout this project, especially Noelle Bathurst, Pat Frederickson, Ryan McAndrews,
Harper Christopher, Jamie Koch, and Matt Diamond.
We would also like to thank Jenna Eisenman, Matthew Hawke, and Megan Hawke for
their outstanding research assistance.
ISTUDY
BRIEF CONTENTS
PART O NE
Introduction
CHAPTER 1 Data and Data Preparation 2
PART TWO
Descriptive Statistics
CHAPTER 2 Tabular and Graphical Methods 32
CHAPER 3 Numerical Descriptive Measures 80
PART TH REE
Probability and Probability Distributions
CHAPTER 4 Introduction to Probability 124
CHAPTER 5 Discrete Probability Distributions 160
CHAPTER 6 Continuous Probability Distributions 200
PART FO UR
Basic Inference
CHAPTER 7 Sampling and Sampling Distributions 238
CHAPTER 8 Interval Estimation 278
CHAPTER 9 Hypothesis Testing 310
CHAPTER 10 Statistical Inference Concerning Two Populations 344
CHAPTER 11 Statistical Inference Concerning Variance 380
CHAPTER 12 Chi-Square Tests 408
PART FI VE
Advanced Inference
CHAPTER 13 Analysis of Variance 438
CHAPTER 14 Regression Analysis 482
CHAPTER 15 Inference with Regression Models 514
CHAPTER 16 Regression Models for Nonlinear Relationships 556
CHAPTER 17 Regression Models with Dummy Variables 588
PART S IX
Supplementary Topics
CHAPTER 18 Forecasting with Time Series Data 622
CHAPTER 19 Returns, Index Numbers, and Inflation 658
CHAPTER 20 Nonparametric Tests 682
A P PENDIXES
APPENDIX A Getting Started with R 721
APPENDIX B Tables 727
APPENDIX C Answers to Selected Even-Numbered Exercises 739
Glossary 755
Index 763
ISTUDY
CONTENTS
xxvi
ISTUDY
4.2 Rules of Probability 133 6.3 Other Continuous Probability Distributions 221
4.3 Contingency Tables and Probabilities 139 The Exponential Distribution 221
A Note on Independence with Empirical Probabilities 141 Using R for the Exponential Distribution 224
The Lognormal Distribution 224
4.4 The Total Probability Rule Using R for the Lognormal Distribution 226
and Bayes’ Theorem 144
The Total Probability Rule and Bayes’ Theorem 144 6.4 Writing with Data 229
Extensions of the Total Probability Rule Conceptual Review 231
and Bayes’ Theorem 146 Additional Exercises 232
4.5 Counting Rules 149 Appendix 6.1: Guidelines for Other Software
4.5 Writing with Data 151 Packages 235
Conceptual Review 154
Additional Exercises 155
PA RT F O U R
CHAPTER 5 Basic Inference
DISCRETE PROBABILITY
CHAPTER 7
DISTRIBUTIONS 160
5.1 Random Variables and Discrete Probability SAMPLING AND SAMPLING
Distributions 162 DISTRIBUTIONS 238
The Discrete Probability Distribution 162
7.1 Sampling 240
5.2 Expected Value, Variance, and Standard
Classic Case of a “Bad” Sample: The Literary Digest
Deviation 166
Debacle of 1936 240
Summary Measures 167
Trump’s Stunning Victory in 2016 241
Risk Neutrality and Risk Aversion 168
Sampling Methods 242
5.3 Portfolio Returns 171 Using Excel and R to Generate a Simple Random Sample 244
Properties of Random Variables 171
7.2 The Sampling Distribution of the Sample Mean 245
Summary Measures for a Portfolio 172
The Expected Value and the Standard Error
5.4 The Binomial Distribution 175 of the Sample Mean 246
Using Excel and R to Obtain Binomial Probabilities 180 Sampling from a Normal Population 247
5.5 The Poisson Distribution 183 The Central Limit Theorem 248
Using Excel and R to Obtain Poisson Probabilities 186 7.3 The Sampling Distribution of the Sample
5.6 The Hypergeometric Distribution 189 Proportion 252
Using Excel and R to Obtain Hypergeometric The Expected Value and the Standard Error
Probabilities 191 of the Sample Proportion 252
5.7 Writing with Data 193 7.4 The Finite Population Correction Factor 257
CONTENTS
ISTUDY
8.2 Confidence Interval For The Population Mean Hypothesis Test for μD 358
When σ Is Unknown 288 Using Excel and R for Testing Hypotheses about μD 361
The t Distribution 288 One Last Note on the Matched-Pairs Experiment 362
Summary of the tdf Distribution 289 10.3 Inference Concerning the Difference Between Two
Locating tdf Values and Probabilities 289 Proportions 366
Constructing a Confidence Interval for μ Confidence Interval for p1 − p2 366
When σ Is Unknown 291 Hypothesis Test for p1 − p2 367
Using Excel and R to Construct a Confidence
10.4 Writing with Data 372
Interval for μ When σ Is Unknown 292
Conceptual Review 374
8.3 Confidence Interval for the Population
Proportion 295 Additional Exercises 375
8.4 Selecting the Required Sample Size 298 Appendix 10.1: Guidelines for Other Software
Selecting n to Estimate μ 299 Packages 377
Selecting n to Estimate p 299
8.5 Writing with Data 302 CHAPTER 11
Conceptual Review 303
Additional Exercises 304 STATISTICAL INFERENCE
Appendix 8.1: Guidelines for Other Software CONCERNING VARIANCE 380
Packages 307 11.1 Inference Concerning
the Population Variance 382
Sampling Distribution of S2 382
CHAPTER 9 Finding χ2dfValues and Probabilities 383
Confidence Interval for the Population Variance 385
HYPOTHESIS TESTING 310 Hypothesis Test for the Population Variance 386
9.1 Introduction to Hypothesis Testing 312 Note on Calculating the p-Value for a Two-Tailed Test
The Decision to “Reject” or “Not Reject” Concerning σ2 387
the Null Hypothesis 312 Using Excel and R to Test σ2 387
Defining the Null and the Alternative Hypotheses 313 11.2 Inference Concerning the Ratio of Two Population
Type I and Type II Errors 315 Variances 391
9.2 Hypothesis Test For The Population Mean Sampling Distribution of S21∕ S22 391
When σ Is Known 318 Finding F (df 1,df 2)Values and Probabilities 392
The p-Value Approach 318 Confidence Interval for the Ratio of Two Population
Confidence Intervals and Two-Tailed Hypothesis Tests 322 Variances 394
One Last Remark 323 Hypothesis Test for the Ratio of Two Population
9.3 Hypothesis Test For The Population Mean Variances 395
When σ Is Unknown 325 Using Excel and R to Test σ21∕ σ22 397
Using Excel and R to Test μ When σ is Unknown 326 11.3 Writing with Data 401
9.4 Hypothesis Test for the Population Conceptual Review 403
Proportion 330 Additional Exercises 403
9.5 Writing with Data 334 Appendix 11.1: Guidelines for Other Software
Conceptual Review 336 Packages 405
Additional Exercises 337
Appendix 9.1: The Critical Value Approach 339 CHAPTER 12
Appendix 9.2: Guidelines for Other Software
Packages 342 CHI-SQUARE TESTS 408
12.1 Goodness-of-Fit Test for
a Multinomial Experiment 410
CHAPTER 10 Using R to Conduct a Goodness-of-Fit Test 414
ISTUDY
CHAPTER 15
PART FIVE
Advanced Inference INFERENCE WITH REGRESSION MODELS 514
15.1 Tests of Significance 516
Test of Joint Significance 516
CHAPTER 13 Test of Individual Significance 518
Using a Confidence Interval to Determine Individual
ANALYSIS OF VARIANCE 438 Significance 520
A Test for a Nonzero Slope Coefficient 521
13.1 One-Way Anova Test 440
Reporting Regression Results 523
Between-Treatments Estimate of σ2: MSTR 441
15.2 A General Test of Linear Restrictions 527
Within-Treatments Estimate of σ2: MSE 442
Using R to Conduct Partial F Tests 530
The One-Way ANOVA Table 444
Using Excel and R to Construct a One-Way ANOVA
15.3 Interval Estimates for the Response Variable 532
Using R to Find Interval Estimates for the Response
Table 444
Variable 535
13.2 Multiple Comparison Methods 449
15.4 Model Assumptions and Common Violations 537
Fisher’s Least Significant Difference (LSD) Method 449 Residual Plots 537
Tukey’s Honestly Significant Difference (HSD) Method 450 Assumption 1. 538
Using R to Construct Tukey Confidence Intervals Detecting Nonlinearities 538
for μ1 − μ2 452
Remedy 539
13.3 Two-Way Anova Test: No Interaction 456 Assumption 2. 539
The Sum of Squares for Factor A, SSA 458 Detecting Multicollinearity 540
The Sum of Squares for Factor B, SSB 459 Remedy 541
The Error Sum of Squares, SSE 459 Assumption 3. 541
Using Excel and R for a Two-Way ANOVA Test—No Detecting Changing Variability 541
Interaction 460 Remedy 542
13.4 Two-Way Anova Test: With Interaction 465 Assumption 4. 543
The Total Sum of Squares, SST 466 Detecting Correlated Observations 543
The Sum of Squares for Factor A, SSA, and the Sum of Remedy 544
Squares for Factor B, SSB 466 Assumption 5. 544
The Sum of Squares for the Interaction of Factor A and Remedy 544
Factor B, SSAB 467 Assumption 6. 545
The Error Sum of Squares, SSE 468 Summary of Regression Modeling 545
Using Excel and R for a Two-Way ANOVA Test—With Using Excel and R for Residual Plots, and R for Robust
Interaction 468 Standard Errors 545
13.5 Writing with Data 472 15.5 Writing with Data 548
Conceptual Review 474 Conceptual Review 550
Additional Exercises 475 Additional Exercises 551
Appendix 13.1: Guidelines for Other Software Packages 479 Appendix 15.1: Guidelines for Other Software Packages 553
CHAPTER 16
CONTENTS
ISTUDY
CHAPTER 17 CHAPTER19
ISTUDY
BUSINESS STATISTICS
ISTUDY
1 Data and Data
Preparation
L E A R N I N G O B J E C T I V E S
After reading this chapter, you should be able to:
I
n just about any contemporary human activity, we use statistics to analyze large amounts of
data for making better decisions. Managers, consumers, sports enthusiasts, politicians, and
medical professionals are increasingly turning to data to boost a company’s revenue, deepen
customer engagement, find better options on consumer products, prevent threats and fraud, suc-
ceed in sports and elections, provide better diagnoses and cures for diseases, and so on. In this
chapter, we will describe various types of data and measurement scales of variables that are used
in statistics.
It is important to note that after obtaining relevant data, we often spend a considerable amount
of time on inspecting and preparing the data for subsequent analysis. In this chapter, we will discuss
a few important data preparation tasks. We will use counting and sorting of relevant variables to
inspect and explore data. Finally, we will discuss a commonly used technique called subsetting,
where only a portion (subset) of the data is used for the analysis.
ISTUDY
Yuliia Mazurkevych/Shutterstoick
I N T R O D U C T O R Y C A S E
ISTUDY
LO 1.1 1.1 TYPES OF DATA
Explain the various types In general, data are compilations of facts, figures, or other contents, both numerical and
of data. nonnumerical. Data of all types and formats are generated from multiple sources. Insights
from all of these data improves a company’s bottom-line and enhances consumer experi-
ence. At the core, business statistics benefits companies by developing better marketing
strategies, deepening customer engagement, enhancing efficiency in procurement, uncov-
ering ways to reduce expenses, identifying emerging market trends, mitigating risk and
fraud, etc. We often find a large amount of data at our disposal. However, we also derive
insights from relatively small data sets, such as from consumer focus groups, marketing
surveys, or reports from government agencies.
Every day, consumers and businesses use data from various sources to help make deci-
sions. In order to make intelligent decisions in a world full of uncertainty, we have to under-
stand statistics—the language of data. In the broadest sense, statistics is the science of
extracting useful information from data. Three steps are essential for doing good statistics.
An important first step for making decisions is to find the right data, which are both com-
plete and lacking any misrepresentation, and prepare it for the analysis. Second, we must
use the appropriate statistical tools, depending on the data at hand. Finally, an important
ingredient of a well-executed statistical analysis is to clearly communicate information into
verbal and written language. It is important to note that numerical results are not very useful
unless they are accompanied with clearly stated actionable business insights.
ISTUDY
Figure 1.1 depicts the flow of information between a population and a sample. Con-
sider, for example, a 2016 Gallop survey that found that only 50% of millennials plan to
stay at their current job for more than a year. Researchers use this sample result, called a
sample statistic, in an attempt to estimate the corresponding unknown population
parameter. In this case, the parameter of interest is the percentage of all millennials who
plan to be with their current job for more than a year.
Sampling
Population Sample
Inference
It is generally not feasible to obtain population data due to prohibitive costs and/or
practicality. We rely on sampling because we are unable to use population data for two
main reasons.
∙ Obtaining information on the entire population is expensive. Consider how the
monthly unemployment rate in the United States is calculated by the Bureau of Labor
Statistics (BLS). Is it reasonable to assume that the BLS counts every unemployed
person each month? The answer is a resounding NO! In order to do this, every home in
the country would have to be contacted. Given that there are approximately 160 million
individuals in the labor force, not only would this process cost too much, it would take an
inordinate amount of time. Instead, the BLS conducts a monthly sample survey of about
60,000 households to measure the extent of unemployment in the United States.
∙ It is impossible to examine every member of the population. Suppose we are
interested in the average length of life of a Duracell AAA battery. If we tested the
duration of each Duracell AAA battery, then in the end, all batteries would be dead
and the answer to the original question would be useless.
ISTUDY
TABLE 1.2 2018–2019 NBA Eastern Conference
Team name Wins Losses Winning percentage
Milwaukee Bucks 60 22 0.732
Toronto Raptors* 58 24 0.707
Philadephia 76ers 51 31 0.622
Boston Celtics 49 33 0.598
Indiana Pacers 48 34 0.585
Brooklyn Nets 42 40 0.512
Orlando Magic 42 40 0.512
Detroit Pistons 41 41 0.500
*The Toronto Raptors won their first NBA title during the 2018–2019 season.
Time series data refer to data collected over several time periods focusing on certain
groups of people, specific events, or objects. Time series data can include hourly, daily,
weekly, monthly, quarterly, or annual observations. Examples of time series data include
the hourly body temperature of a patient in a hospital’s intensive care unit, the daily price
of General Electric stock in the first quarter of 2020, the weekly exchange rate between
the U.S. dollar and the euro over the past six months, the monthly sales of cars at a dealer-
ship in 2020, and the annual population growth rate of India in the last decade. In these
examples, temporal ordering is relevant and meaningful.
Figure 1.2 shows a plot of the national homeownership rate in the U.S. from 2000 to
2018. According to the U.S. Census Bureau, the national homeownership rate in the first
quarter of 2016 plummeted to 63.6% from a high of 69.4% in 2004. An explanation for the
decline in the homeownership rate is the stricter lending practices caused by the housing
market crash in 2007 that precipitated a banking crisis and deep recession. This decline
can also be attributed to home prices outpacing wages in the sample period.
68.0
66.0
64.0
62.0
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Year
ISTUDY
Examples of structured data include numbers, dates, and groups of words and numbers,
typically stored in a tabular format. Structured data often consist of numerical informa-
tion that is objective and is not open to interpretation.
Point-of-sale and financial data are examples of structured data and are usually
designed to capture a business process or transaction. Examples include the sale of retail
products, money transfer between bank accounts, and the student enrollment in a univer-
sity course. When individual consumers buy products from a retail store, each transaction
is captured into a record of structured data.
Consider the sales invoice shown in Figure 1.3. Whenever a customer places an order
like this, there is a predefined set of data to be collected, such as the transaction date,
shipping address, and the units of product being purchased. Even though a receipt or an
invoice may not always be presented in rows and columns, the predefined structure
allows businesses and organizations to translate the data on the document into a
row-column format.
For decades, companies and organizations relied mostly on structured data to run
their businesses and operations. Today, with the advent of the digital age, most experts
agree that only about 20% of all data used in business decisions are structured data. The
remaining 80% are unstructured.
Unlike structured data, unstructured data (or unmodeled data) do not conform to a
predefined, row-column format. They tend to be textual (e.g., written reports, e-mail mes-
sages, doctor’s notes, or open-ended survey responses) or have multimedia contents (e.g.,
photographs, videos, and audio data). Even though these data may have some implied
structure (e.g., a report title, e-mail’s subject line, or a time stamp on a photograph), they
are still considered unstructured as they do not conform to a row-column model required
in most database systems. Social media data such as Twitter, YouTube, Facebook, and
blogs are examples of unstructured data.
Big Data
Nowadays, businesses and organizations generate and gather more and more data at an
increasing pace. The term big data is a catch-phrase, meaning a massive amount of both
structured and unstructured data that are extremely difficult to manage, process, and
ISTUDY
analyze using traditional data-processing tools. Despite the challenges, big data present
great opportunities to gain knowledge and business intelligence with potential game-
changing impacts on company revenues, competitive advantage, and organizational
efficiency.
More formally, a widely accepted definition of big data is “high-volume, high-velocity
and/or high-variety information assets that demand cost-effective, innovative forms of
information processing that enable enhanced insight, decision making, and process
automation” (www.gartner.com). The three characteristics (the three Vs) of big data are:
∙ Volume: An immense amount of data is compiled from a single source or a wide
range of sources, including business transactions, household and personal devices,
manufacturing equipment, social media, and other online portals.
∙ Velocity: In addition to volume, data from a variety of sources get generated at a rapid
speed. Managing these data streams can become a critical issue for many organizations.
∙ Variety: Data also come in all types, forms, and granularity, both structured and
unstructured. These data may include numbers, text, and figures as well as audio,
video, e-mails, and other multimedia elements.
In addition to the three defining characteristics of big data, we also need to pay close
attention to the veracity of the data and the business value that they can generate. Veracity
refers to the credibility and quality of data. One must verify the reliability and accuracy
of the data content prior to relying on the data to make decisions. This becomes
increasingly challenging with the rapid growth of data volume fueled by social media and
automatic data collection. Value derived from big data is perhaps the most important
aspect of any statistical project. Having a plethora of data does not guarantee that useful
insights or measurable improvements will be generated. Organizations must develop a
methodical plan for formulating business questions, curating the right data, and unlocking
the hidden potential in big data.
Big data, however, do not necessarily imply complete (population) data. Take, for
example, the analysis of all Facebook users. It certainly involves big data, but if we
consider all Internet users in the world, Facebook users are only a very large sample.
There are many Internet users who do not use Facebook, so the data on Facebook do not
represent the population. Even if we define the population as pertaining to those who use
online social media, Facebook is still one of many social media portals that consumers
use. And because different social media are used for different purposes, data collected
from these sites may very well reflect different populations of Internet users; this
distinction is especially important from a strategic business standpoint. Therefore,
Facebook data are simply a very large sample.
In addition, we may choose not to use big data in its entirety even when they are avail-
able. Sometimes it is just inconvenient to analyze a very large data set as it is computa-
tionally burdensome, even with a modern, high-capacity computer system. Other times,
the additional benefits of working with big data may not justify the associated costs. In
sum, we often choose to work with relatively smaller data sets drawn from big data.
ISTUDY
Data on the Web
The explosion in the field of statistics and data analytics is partly due to the
growing availability of vast amounts of data and improved computational power.
Many experts believe that 90% of the data in the world today were created in the
last two years alone. These days, it has become easy to access data by simply
using a search engine like Google. These search engines direct us to data-
providing websites. For instance, searching for economic data may lead you to
the Bureau of Economic Analysis (www.bea.gov), the Bureau of Labor Statistics
(www.bls.gov/data), the Federal Reserve Economic Data (research.stlouisfed.
org), and the U.S. Census Bureau (www.census.gov/data.html). These websites
provide data on inflation, unemployment, GDP, and much more, including useful
international data.
The National Climatic Data Center (www.ncdc.noaa.gov/data-access) pro-
vides a large collection of environmental, meteorological, and climate data. Sim-
ilarly, transportation data can be found at www.its-rde.net. The University of
Michigan has compiled sentiment data found at www.sca.isr.umich.edu. Several
cities in the United States have publicly available data in categories such as
finance, community and economic development, education, and crime. For
example, the Chicago data portal data.cityofchicago.org provides a large volume Comstock Images/Jupiterimages
of city-specific data. Excellent world development indicator data are available at
data.worldbank.org. The happiness index data for most countries are available at www.
happyplanetindex.org/data.
Private corporations also make data available on their websites. For example, Yahoo
Finance (www.finance.yahoo.com) and Google Finance (www.google.com/finance) list
data such as stock prices, mutual fund performance, and international market data. Zillow
(www.zillow.com/) supplies data for recent home sales, monthly rent, mortgage rates, and
so forth. Similarly, www.espn.go.com offers comprehensive sports data on both profes-
sional and college teams. Finally, The Wall Street Journal, The New York Times, USA
Today, The Economist, Business Week, Forbes, and Fortune are all reputable publications
that provide all sorts of data. We would like to point out that all of the above data sources
represent only a fraction of publicly available data.
E X E RC I S E S 1.1
Applications computes the average GPA as 3.29. Describe the relevant
population.
1. A few years ago, it came as a surprise when Apple’s iPhone
4 was found to have a problem. Users complained of weak 4. Recent college graduates with an engineering degree continue
reception, and sometimes even dropped calls, when they to earn high salaries. An online search revealed that the average
cradled the phone in their hands in a particular way. A survey annual salary for an entry-level position in engineering is $65,000.
at a local store found that 2% of iPhone 4 users experienced a. What is the relevant population?
this reception problem. b. Do you think the average salary of $65,000 is computed
a. Describe the relevant population. from the population? Explain.
b. Is 2% associated with the population or the sample? 5. Research suggests that depression significantly increases the
2. Many people regard video games as an obsession for young- risk of developing dementia later in life. Suppose that in a
sters, but, in fact, the average age of a video game player is study involving 949 elderly persons, it was found that 22% of
35 years old. Is the value 35 likely the actual or the estimated those who had depression went on to develop dementia, com-
average age of the population? Explain. pared to only 17% of those who did not have depression.
3. An accounting professor wants to know the average GPA of a. Describe the relevant population and the sample.
the students enrolled in her class. She looks up information b. Are the numbers 22% and 17% associated with the
on Blackboard about the students enrolled in her class and population or a sample?
ISTUDY
6. Go to www.zillow.com and find the sale price of 20 single-family 9. Conduct an online search to compare small hybrid vehicles
homes sold in Las Vegas, Nevada, in the last 30 days. Structure (e.g., Toyota Prius, Ford Fusion, Chevrolet Volt) on price, fuel
the data in a tabular format and include the sale price, the num- economy, and other specifications. Do you consider the search
ber of bedrooms, the square footage, and the age of the house. results structured or unstructured data? Explain.
Do these data represent cross-sectional or time series data? 10. Find Under Armour’s annual revenue from the past 10 years.
7. Go to www.finance.yahoo.com to get the current stock quote Are the data considered structured or unstructured? Explain.
for Home Depot (ticker symbol = HD). Use the ticker symbol to Are they cross-sectional or time series data?
search for historical prices and create a table that includes the 11. Ask 20 of your friends about their online social media usage,
monthly adjusted close price of Home Depot stock for the last specifically whether or not they use Facebook, Instagram,
12 months. Do these data represent cross-sectional or time and Snapchat; how often they use each social media portal;
series data? and their overall satisfaction of each of these portals. Create
8. Go to The New York Times website at www.nytimes.com and a table that presents this information. Are the data considered
review the front page. Would you consider the data on the structured or unstructured? Are they cross-sectional or time
page to be structured or unstructured? Explain. series data?
ISTUDY
CATEGORICAL AN D NUMERICAL VARIABLES
A variable is a general characteristic being observed on a set of people, objects, or
events, where each observation varies in kind or degree.
∙ The observations of a categorical variable assume names or labels.
∙ The observations of a numerical variable assume meaningful numerical val-
ues. A numerical variable can be further categorized as either discrete or con-
tinuous. A discrete variable assumes a countable number of values, whereas a
continuous variable is characterized by uncountable values within an interval.
EXAMPLE 1.1
In the introductory case, Catherine Hill has been assigned to help market Organic
Food Superstore’s new line of Asian-inspired meals. With the help of the IT group,
she has acquired a representative sample of customers at her store. Determine
which of the variables in the sample are categorical or numerical and, if numerical,
determine if they are discrete or continuous.
SOLUTION:
The variables Sex, Race, College, and Channel are categorical, merely representing
labels. We also treat Birthdate as a categorical variable, with numerous categories,
even though it contains numbers. Note that we can easily convert date of birth to a
numerical variable age by simply subtracting it from the current date. On the other
hand, HHSize, Income, Spending, and Orders are numerical variables because the
observations are all meaningful numbers. Note that all of the numerical variables in
this example are discrete because they can only assume a countable number of values;
in other words, they are not characterized by uncountable values within an interval.
ISTUDY
TABLE 1.3 Companies of the DJIA and Exchange Where Stock is Traded
Company Exchange Company Exchange
3M (MMM) NYSE Johnson & Johnson (JNJ) NYSE
American Express (AXP) NYSE JPMorgan Chase (JPM) NYSE
Apple (AAPL) Nasdaq McDonald’s (MCD) NYSE
Boeing (BA) NYSE Merck (MRK) NYSE
Caterpillar (CAT) NYSE Microsoft (MFST) Nasdaq
Chevron (CVX) NYSE Nike (NKE) NYSE
Cisco (CSCO) Nasdaq Pfizer (PFE) NYSE
Coca-Cola (KO) NYSE Procter & Gamble (PG) NYSE
Disney (DIS) NYSE Travelers (TRV) NYSE
DowDupont (DWDP) NYSE United Health (UNH) NYSE
ExxonMobil (XOM) NYSE United Technologies (UTX) NYSE
Goldman Sachs (GS) NYSE Verizon (VZ) NYSE
Home Depot (HD) NYSE Visa (V) NYSE
IBM (IBM) NYSE Wal-Mart (WMT) NYSE
Intel (INTC) Nasdaq Walgreen (WBA) Nasdaq
Compared to the nominal scale, the ordinal scale reflects a stronger level of measure-
ment. With ordinal observations, we are able to both categorize and rank them with
respect to some characteristic or trait. The weakness with ordinal observations is that we
cannot interpret the difference between the ranked observations because the actual num-
bers used are arbitrary. Consider, for example, hotel reviews where consumers are asked
to classify the service at a particular hotel as excellent (5 stars), very good (4 stars), good
(3 stars), fair (2 stars), or poor (1 star). We summarize the categories and their respective
ratings in Table 1.4.
TABLE 1.4 Hotel Survey
Categories with Ratings
Category Rating
Excellent 5
Very good 4
Good 3
Fair 2
Poor 1
In Table 1.4, the number attached to excellent (5 stars) is higher than the number attached
to good (3 stars), indicating that the response of excellent is preferred to good. However,
we can easily redefine the ratings, as we show in Table 1.5.
Category Rating
Excellent 100
Very good 80
Good 70
Fair 50
Poor 40
ISTUDY
In Table 1.5, excellent still receives a higher number than good, but now the difference
between the two categories is 30 points (100 – 70), as compared to a difference of 2
points (5 – 3) when we use the first classification. In other words, differences between
categories are meaningless with ordinal observations. (We also should note that we could
reverse the ordering so that, for instance, excellent equals 40 and poor equals 100; this
renumbering would not change the nature of the observations.)
As mentioned earlier, observations of a categorical variable are typically expressed in
words but are coded into numbers for purposes of data processing. When summarizing
the results of a categorical variable, we typically count the number of observations that
fall into each category or calculate the percentage of observations that fall into each cat-
egory. However, with a categorical variable, we are unable to perform meaningful arith-
metic operations, such as addition and subtraction.
With observations that are measured on the interval scale, we are able to categorize
and rank them as well as find meaningful differences between them. The Fahrenheit scale
for temperatures is an example of interval-scaled variable. Not only is 60 degrees Fahren-
heit hotter than 50 degrees Fahrenheit, the same difference of 10 degrees also exists
between 90 and 80 degrees Fahrenheit.
The main drawback of an interval-scaled variable is that the value of zero is
arbitrarily chosen; the zero point of an interval-scaled variable does not reflect a
complete absence of what is being measured. No specific meaning is attached to
0 degrees Fahrenheit other than to say it is 10 degrees colder than 10 degrees Fahrenheit.
With an arbitrary zero point, meaningful ratios cannot be constructed. For instance, it
is senseless to say that 80 degrees is twice as hot as 40 degrees; in other words, the ratio
80/40 has no meaning.
The ratio scale represents the strongest level of measurement. The ratio scale has all
the characteristics of the interval scale as well as a true zero point, which allows us to
interpret the ratios between observations. The ratio scale is used in many business appli-
cations. Variables such as sales, profits, and inventory levels are expressed on the ratio
scale. A meaningful zero point allows us to state, for example, that profits for firm A are
double those of firm B. Variables such as weight, time, and distance are also measured on
a ratio scale because zero is meaningful.
Unlike nominal- and ordinal-scaled variables (categorical variables), arithmetic oper-
ations are valid on interval- and ratio-scaled variables (numerical variables). In later
chapters, we will calculate summary measures, such as the mean, the median, and
the variance, for numerical variables; we cannot calculate these measures for categorical
variables.
MEASUREMENT SCALES
The observations for any variable can be classified into one of four major measure-
ment scales: nominal, ordinal, interval, or ratio.
∙ Nominal: Observations differ merely by name or label.
∙ Ordinal: Observations can be categorized and ranked; however, differences
between the ranked observations are meaningless.
∙ Interval: Observations can be categorized and ranked, and differences between
observations are meaningful. The main drawback of the interval scale is that
the value of zero is arbitrarily chosen.
∙ Ratio: Observations have all the characteristics of an interval-scaled variable
as well as a true zero point; thus, meaningful ratios can be calculated.
Nominal and ordinal scales are used for categorical variables, whereas interval and
ratio scales are used for numerical variables.
ISTUDY
EXAMPLE 1.2
The owner of a ski resort two hours outside Boston, Massachusetts, is interested
Tween_Survey in serving the needs of the “tween” population (children aged 8 to 12 years old).
He believes that tween spending power has grown over the past few years, and he
wants their skiing experience to be memorable so that they want to return. At the
end of last year’s ski season, he asked 20 tweens the following four questions.
∙ Q1. On your car drive to the resort, which music streaming service was
playing?
∙ Q2. On a scale of 1 to 4, rate the quality of the food at the resort (where 1 is
poor, 2 is fair, 3 is good, and 4 is excellent).
∙ Q3. Presently, the main dining area closes at 3:00 pm. What time do you think
it should close?
∙ Q4. How much of your own money did you spend at the lodge today?
A portion of their responses is shown in Table 1.6. Identify the scale of measure-
ment for each variable used in the survey. Given the tween responses, provide sug-
gestions to the owner for improvement.
SOLUTION:
∙ Q1. Responses for music streaming service are nominal because the observa-
tions differ merely in label. Twelve of the 20 tweens, or 60%, listened to Spo-
tify. If the resort wishes to contact tweens using this means, then it may want
to direct its advertising dollars to this streaming service.
∙ Q2. Food quality responses are on an ordinal scale because we can both
categorize and rank the observations. Eleven of the 20 tweens, or 55%, felt
that the food quality was, at best, fair. Perhaps a more extensive survey that
focuses solely on food quality would reveal the reason for their apparent
dissatisfaction.
∙ Q3. Closing time responses are on an interval scale. We can say that 3:30 pm
is 30 minutes later than 3:00 pm, and 6:00 pm is 30 minutes later than 5:30 pm;
that is, differences between observations are meaningful. The closing time
responses, however, have no apparent zero point. We could arbitrarily define
the zero point at 12:00 am, but ratios are still meaningless. In other words,
it makes no sense to form the ratio 6:00 pm/3:00 pm and conclude that
6:00 pm is twice as long a time period as 3:00 pm. A review of the closing
time responses shows that the vast majority (19 out of 20) would like the
dining area to remain open later.
∙ Q4. The tweens’ responses with respect to their own money spent at the resort
are on a ratio scale. We can categorize and rank observations as well as calculate
meaningful differences. Moreover, because there is a natural zero point, valid
ratios can also be calculated. Seventeen of the 20 tweens spent their own money at
the lodge. It does appear that the discretionary spending of this age group is sig-
nificant. The owner would be wise to cater to some of their preferences.
ISTUDY
Another random document with
no related content on Scribd:
lahonneen kannon kumoon tahi ottaa hengiltä mahdollisesti eteen
sattuvan kirjavan kyyn.
Mitkä?
Hyljetäit!
Pitkin käsiä ja kaulaa niitä matelee ja ne purevat, purevat kuin
vimmatut, polttavat kuin tuli! Hilpeän pirullisella mielellä koetan
huvitella itseäni katselemalla niitä suurennuslasini läpi, muistelen
että ne ovat Haematopinus nimisiä — mutta lopputulos on vain
kaiken muun voittava tunne — ne purevat!
*****
Päinvastoin!
Vaan — kärsivällisyyttä!