Visual Revelations Graphical Tales of Fate and Deception

From Napoleon Bonaparte To Ross Perot, 1st Edition

Visit the link below to download the full version of this book:

https://medipdf.com/product/visual-revelations-graphical-tales-of-fate-and-decep
tion-from-napoleon-bonaparte-to-ross-perot-1st-edition/

Click Download Now

 Originally published in 1997.

© 1997 Howard Wainer

All rights reserved. No part of this book may be reproduced in

any form, by photostat, microform, retrieval system, or any other
means, without the prior written permission of the publisher.

Lawrence Erlbaum Associates, Inc.

10 Industrial A venue
Mahwah, New Jersey 07430

About the cover: "Chart of the National Debt of England," which appeared as
plate 20, opposite p. 83 in William Playfair's CommerciaL and PoLitical Atlas
(third edition), published in London in 1801. This beautifully executed line chart
uses the innovation of an irregularly spaced grid along the time (horizontal) axis
to demark events of important economic consequence. The inexorable conclusion
we draw is that war was bad for England's national debt.

Library of Congress Cataloging-in-Publication Data

Wainer, Howard.
Visual revelations: graphical tales of fate and deception from
Napoleon Bonaparte to Ross Perot / Howard Wainer.
p. cm.
Originally published: New York: Copernicus, 1997.
Includes bibliographical references and index.
ISBN 0-8058-3878-3 (cloth: alk. paper)
I. Mathematical statistics-Graphic methods. I. Title.

QA276.3.W35 2000
001.4'226-dc21
00-056185

1098765432 I

Copyrighted Material
To Linda and Sam

Copyrighted Material
This page intentionally left blank

Copyrighted Material
 Contents

.. -

Acknowledgments 1X

Introduction 1
SECTION I Graphical Failures 5
CHAPTER 1 How to Display Data Badly 11
CHAPTER 2 Graphical Mysteries 47

SECTION II Graphical Triumphs 55

CHAPTER 3 Graphical Answers to 57
Scientific Q!lestions
CHAPTER 4 Three Graphic Memorials 63
CHAPTER 5 A Nobel Graph 69
CHAPTER 6 Todai Moto Kurashi 75
CHAPTER 7 Picturing an L.A. Bus Schedule 79

SECTION III Graphical Forms 85

CHAPTER 8 Humble Pie 87
CHAPTER 9 Double Y-Axis Graphs 91
CHAPTER 10 Tabular Presentation 95
CHAPTER 11 A Rose by Another Name 103
CHAPTER 12 Trilinear Plots 111
CHAPTER 13 Implicit Graphs 119

[ VII]

Copyrighted Material
[ VIII] CONTENTS

SECTION IV Using Graphical Methods 125

CHAPTER 14 Measuring Graphicacy 127
CHAPTER 15 Graphs in the Presidential 129
Campaign: Why Weren't They
Used More Broadly?
CHAPTER 16 Visual Aids When Comparing 134
an Apple to the Stars

SECTION V Improving Graphical Presentations 141

CHAPTER 17 Integrating Figures and Text 143
CHAPTER 18 Elegance, Grace, Impact, 146
and Graphical Displays
CHAPTER 19 Sense-Lining 151
CHAPTER 20 Making Readable 155
Overhead Displays
Finally- 161

Notes 163
References 166
Credits 172
Index 173

Copyrighted Material
 Acknowledgments

- .-
The writing of this book spans almost twenty-five years. In that time
an enormous number of intellectual debts have accumulated. I am de-
lighted to be able to acknowledge my gratitude to the many who have
contributed to my interest in and understanding of graphics.
The path began with John W. Tukey, who asked me to act as a dis-
cussant at a symposium on exploratory data analysis he had organized
in 1972. Feeling that discussing the statistical technology of the fore-
most leaders in that field was an overreach for someone only barely
out of graduate school, I decided to center my discussion on the need
to verifY the efficacy of graphical innovations experimentally. Fred
Mosteller, who was also part of the symposium, suggested that I pub-
lish my findings. I did, and thus my graphical career began. That arti-
cle came to the attention of Albert Biderman, who was looking for
someone to run a research project at the Bureau of Social Science
Research on the uses of graphics for social reporting. For many rea-
sons this seemed like an interesting path to follow, so I agreed enthu-
siastically to join, and moved to Washington. One product of this
project was the translation ofJacques Bertin's monumental Semiologie
Graphique into English, on which Bill Berg and I collaborated hap-
pily. The ideas I absorbed from Bertin and Tukey remain two of the
principal legs supporting the seat of my graphical experience.
The third leg is Edward Tufte. I first met Edward when he gave a
lecture on improving data display at the University of Chicago in 1977.
Although his thinking on graphics was still at an early stage, his ideas
had a profound influence on me; figures 8, 9, and 10 in chapter 1 all
came from that 1977 lecture. Many of the ideas and examples in this
book (especially chapter 1) have their origins with Tufte. I am grateful
indeed for his permission to reprint his figures. There is no one whose
work I admire more than Tufte's. His three books (1983, 1990, 1997)
are a wonderful and beautiful manifestation of his thoughts and taste.
During the course of the Graphics Social Reporting Project
(1977-1980), I met with many of the principal workers in statistical

Copyrighted Material
[x] ACKNOWLEDGMENTS

graphics, but my friend and colleague Albert Biderman was, and re-
mains, enormously influential. Indeed, some of the prose in the intro-
duction to section I as well as the introduction to chapter 5 had its
origin in a piece that we cowrote; I am pleased that Al has given his
permission to reuse it here. Many of the ideas presented in this book
are my interpretations of his Delphic suggestions. Of course, he
should not be held responsible for how I've used them.
Since 1980 I have been employed by the Educational Testing
Service. Although only a little of my work there is directly related to
data display, I have always been accorded time and support to pursue
my graphical interests. Without the forbearance of my employer this
book would never have been completed, and to that I am indebted. I
would specifically like to express my appreciation to the trustees of
the Educational Testing Service, who through the 1990 Senior
Scientist Award afforded me additional freedom to pursue various
graphical notions. In the preface to his Sketches from a Lift, the storied
diplomat and historian George Kennan quoted Anton Chekov, a doc-
tor, who wrote that while medicine was his wife, literature was his
mistress. Kennan added that his situation was similar, but his mistress
was neither as beautiful nor as varied as Chekov's. My relationship
with statistics and graphics follows in the same pattern, although
trailing far behind in breadth and depth as well as beauty.
In many books one often sees statements like "To Zelda, my long-
suffering wife, for late dinners and her understanding that following
my muse precludes doing the dishes." Well, I would like to thank my
wife, Linda Steinberg, for her intolerance. She has read everything in
here, as well as much that happily is not. She has been intolerant of
pompous or unclear prose, of sloppy thinking, and of pointless discus-
sion. Her contribution, in short, was not to the infrastructure that al-
lowed me to write this book, but rather to its intellectual content. My
gratitude, which is immense, is for her direct contributions toward
making this a better book. Any late dinners I usually ended up cook-
ing myself.
Over long periods of time, a number of my colleagues have pro-
vided bits and pieces of help, which when aggregated have become
substantial. Principal among these is my former student, long-term
friend, and frequent collaborator, David Thissen. We have written so
much together that it is often impossible to determine on whose word
processor the prose, and indeed the ideas, originated. I believe that the
prose in this book is mostly mine, but some of the ideas are surely
David's. Happily, he has so many ideas that he doesn't begrudge my
borrowing a few.
In addition, I would like to thank for their comments, sugges-
tions, and contributions to various pieces Jill Callahan, Jeff Douglas,
John Durso, Andrew Ehrenberg, Molson Export, Lawrence Frase,

Copyrighted Material
 ACKNOWLEDGMENTS

Martin Gilchrist, Charles Lewis, Alan MacEachren, John Mazzeo,

Bob Mislevy, Carol Myford, Jim Ramsay, John Rolph, Mary Vaianfl,
Steven Wang, Warren Willingham, and Denise Woerner. I would
also like to thank Patricia Clare Haskell for advice of many sorts; also
for letting me keep in most of the semicolons.
There are many staff members at Springer-Verlag who have had an
important role in bringing this book to market in its current form .
Principal among them are Steven Pisano (production) whose good
humor and sharp eye were instrumental in producing a work that was
as good as he could make it; Karen Philips (designer) who not only
produced the cover but also took seriously my sometime amateurish
suggestions; and Martin Gilchrist (editor) whose enthusiasm for the
project and the subject brought me to Springer in the first place.
Last, it is important to note the genesis of much of this book's con-
tents. Chapter 1 is an expansion of a 1984 article with the same title
that appeared in the American Statistician. Chapters 2 through 20 are
variations on columns I wrote in the statistics magazine Chance over
the time period 1990 through 1996.

Howard Wainer
Princeton, New Jersey
April, 1997

Copyrighted Material
This page intentionally left blank

Copyrighted Material
 Introduction

'Paleolithic cave art provides an

E
early and very striking example
ffectively conveying information in words is a difficult task. of graphic display. Some Ice Age
When the information is complex, the task often becomes bone carvings of animals are in-
insuperable. Attempts to overcome this difficulty using non- termixed with patterns of dots
and strokes that archeologists
linguistic means date back to preclassical antiquity.* The continuing have interpreted as a lunar nota-
search for a more complete answer took a positive turn around 3800 tion system related to the ani-
B.C. when ancient Egyptian geographers transformed spatial infor- mals' seasonal appearance. These
are almost identical in structure,
mation into spatial diagrams.t Since there was no change in the as well as degree of detail, to the
metaphor of display, this was not a huge intellectual leap; but imagine engraving on the hull of the
how much worse off we would be if they had instead transformed the Pioneer 10 Spacecraft, which
shows a drawing of a man and a
same spatial information into words. woman along with a simple
A real breakthrough in conveying quantitative information took plotting of the Earth's location
place in eighteenth-century Britain, when William Playfair (1759- by dotted pulsar beams.
1823), the ne'er-do-well younger brother of the well-known scientist tSpatial diagrams are of course
John Playfair** and a draftsman employed by James Watt, extended maps. The earliest maps to sur-
vive were from Mesopotamia
the graphical metaphor to nonspatial data. He invented many of the and were developed to keep track
currently popular graphical forms, improved the few that already ex- of land shifted by river floods .
isted, and broadly popularized the idea of graphic depiction of quan- "John Playfair's activities were
titative information. Before Playfair, the use of statistical graphics was remarkably varied: minister,
geologist, mathematician, and
narrowly employed and even more narrowly circulated (see chapter 2). professor of natural philosophy
But afterwards graphs popped up everywhere, being used to convey at Edinburgh University. In
economic information, but also for such varied purposes as showing fact, in 1805 William thanked
his brother for the idea of us-
the distribution of ignorance in England and the frequency of im- ing "lines applied to matters of
provident marriages in Wales.! finance" that William used in
By the end of the nineteenth century, data-based graphics were his 1786 book. We can only
speculate why it took him
widely used, and most of the graphical arrows that fill the modern de- nineteen years to give his
signer's quiver had been invented. Indeed, the nineteenth century is an brother some credit.

[ 1]

Copyrighted Material
[ 2] INTRODUCTION

important contributor to the pantheon of the most wonderful graphics

ever produced. But the twentieth century has also made important
contributions to the theory and practice of graphical display. These
contributions have been driven by the critical need for decision-mak-
ers to be able to absorb huge amounts of information. Such a need was
explicitly stated by Playfair in 1786 as the driving force behind his in-
ventions, but in the intervening two centuries, the growth of data-
gathering has been extraordinary. Associated with an abundance of
data is the ubiquity of the computer to manipulate those data and var-
ious kinds of output devices that can display them. Pen nibs and ac-
counting books have been replaced by color laser printers and
CD-ROMs. Static plots are now augmented by dynamic multicolored
displays that were impractical without modern high-speed computing.
The widespread understanding of how to display information ef-
fectively has lagged behind the technology for gathering, manipulat-
ing, and displaying that information. Happily, there have been a
number of truly exemplary discussions of data displays,2 which for the
most part have been scholarly, serious, precise, thorough, and, if the
truth be told, a little dull. The dullness is surprising since the design
of charts and diagrams, representing, as it does, a synthesis of art and
science, has the potential for carrying the glory of each. With the no-
table exception of Edward Tufte's remarkable work3 there have been
too few celebrations of that glory. Visual Revelations is my attempt to
help remedy that lack.
In the subtitle of this book is the word deception, which deserves
some explanation. When we see a chart or diagram, we generally inter-
pret its appearance as a sincere desire on the part of the author to in-
form. In the face of this sincerity, the misuse of graphical material is a
perversion of communication, equivalent to putting up a detour sign
that leads to an abyss. The first section of this book examines the char-
acter of such immoral behavior. The seriousness of the sin committed
by chartists who deceive is multiplied when it is viewed in the context
of how powerful a tool for communication graphical methods can be.
In section II, on the other hand, are reproduced some very wonderful
successes. Napoleon Bonaparte appears in this section (chapter 4),
although Ross Perot doesn't show up until much later (chapter 15).
Between the two, you will meet some unfamiliar characters like the
nineteenth-century French straphanger E.J. Marey, who provides
help to anyone who has ever tried to figure out a train schedule, as
well as familiar figures like Florence Nightingale, who appears here in
an unfamiliar role.
This book can be read as a series of episodes; one can jump in and
read any chapter, and it should, more or less, stand on its own, but the
linear order of their presentation does add a little bit of useful struc-
ture. There is no ponderous theoretical structure that underlies the

Copyrighted Material
 INTRODUCTION [ 3]

advice given. But there is a coherent message: revelation accompanies

simplicity. This should not be confused to mean that successful dis-
plays must be very limited in content. Actually, quite the opposite is
true. We must always measure the grace and simplicity of the display
relative to the complexity of its message. Thus, when two charts con-
tain the same information, the simpler will also be the more revealing.
Graphic display achieves its highest goals when it allows access to im-
portant, complex information; when, in John Tukey's words, "it forces
us to see what we never expected."
It is my hope that after traveling down the path of these visual rev-
elations, your standards will change: your appreciation of the elegance
of a good display will be increased, your tolerance for poor displays
will be decreased, and your ability to tell the difference will be sharp-
ened. I also hope that you will enjoy the journey.

Copyrighted Material
This page intentionally left blank

Copyrighted Material
 SECTION I

Graphical Failures

"0 urforefathers," wrote Shakespeare, in Henry VI, Part II (Act IV),

"had no other books but the score and the tally. " But both score
and tally represented an advance from more literally iconic
predecessors to the concept of abstract number. The recently discov-
ered Combe d'Arc cave drawings in France indicated that ancient
shepherds kept track of the size of their flocks through accurate draw-
ings of them. This work was sufficiently tedious so that it was re-
placed by a simplified, idealized form. This simplification was later
replaced by Shakespeare's tally (111111) and eventually by the abstract
form "6 sheep." The movement from literal to abstract representation
of information traces the history of graphic display. Note that one re-
sult of this movement is that our use of the term graphic has come to
mean the opposite of graphic in the sense of literal Iifelikeness.
Human beings have been using iconic representations such as
maps for millennia. The spatial representation of space is an obvious
metaphor. But the spatial representation of nonspatial data-such as
showing rising and falling imports over time as a rising and falling
line-is both more recent and a deeper intellectual accomplishment.
This major conceptual breakthrough in graphical presentation
came in 1786 with the publication of William Playfair's Political
Atlas, in which spatial dimensions were used to represent nonspatial,
quantitative, idiographic, empirical data. Such a representation now
seems natural, but before that time it was rarely done and was hence
quite an accomplishment. Notably, in addition to the statistical line
chart, Playfair at one fell swoop single-handedly invented most of the
remaining forms of the statistical graphic repertoire used today-the

[ 5]

Copyrighted Material
[6] SECTION I. GRAPHICAL FAILURES

bar chart and histogram, the surface chart, and the circle diagram, or
"pie chart."
I commemorated the bicentennial anniversary of the publication of
Playfair's Atlas by visiting the corporate offices of a tobacco company
(now a tobacco and baked goods company) to give a talk on effective
data display. In preparation I searched for an effective display on a
topic that I thought would be of interest to them. I found one in the
Surgeon General's Report on Smoking.l

Death Rate (Log Scale) Plotted Against Age

Prospective Study of Mortality in U.S. Veterans
• Smokers
7

---'"..
....,
Non·Smokers
<=
.: 6
<=
e
Q
Q

.....
:- 5

.
Q.,

FIGURE 1. The mortality rates

2+---~----.---~---.----~---r--~----.
for smokers and nonsmokers
at various ages shown in a 40 50 60 70 80
Age
log scale.

In figure 1 we see clearly that no matter how old you are, you are
likely to die sooner if you smoke. I thought it effectively summarized
a great deal of information. My audience unanimously disagreed.
Later that afternoon, as I spent some time with my host discussing
these surprising objections, I began to understand. I sketched what I
thought they might consider to be a more effective alternative (figure 2).
By replacing two lines with stacked three-dimensional bars, the obser-
vation of smoking's danger, so clear in the Surgeon General's graph,
becomes considerably more difficult.
"Much better," my host said, "but the dark bars that represent smok-
ers are still visibly longer than the white ones. See if you can do better."
I tried some other alternatives, including labeling the plot in Greek
(figure 3), which drew an appreciative clap on the shoulder and the
encouragement to keep trying.

Copyrighted Material