Hypothesis Follow Up #3

The Fundamental Principles of Analytical Design

Categories such as time, space, cause, and number represent the most general

relations which exist between things; surpassing all our other ideas in extension

they dominate all the details of our intellectual life. If humankind did not

agree upon these essential ideas at every moment, if they did not have the same

conception of time, space, cause, and number, all contact between their minds

would be impossible. . . .

Emile Durkheim, Les formes élémentaires de la vie religieuse

(Paris, 1912), 22-23.

Ido not paint things, I paint only the differences between things.

Henri Matisse, Henri Matisse Dessins: themes et variations

(Paris, 1943), 37.

EXcELLENT graphics exemplify the deep fundamental principles of analytical design in action. If this were not the case, then something might well be wrong with the principles.

Charles Joseph Minard’s data~map describes the successive losses in men of the French army during the French invasion of Russia in 1812. Vivid historical content and brilliant design combine to make this one of the best statistical graphics ever. Carefully study Minard’s graphic, shown in our English translation at right. A title announces the design method (figurative map) and subject (what befell the French army in Russia). Minard identifies himself and provides a credential. A paragraph of text explains the color code, the 3 scales of measurement, and the methods. Five data sources are acknowledged.

Then the Russian campaign of 1812-1813 begins.

At right, English translation of item 28 in Charles Joseph Minard, Tableaux Graphiques

et Cartes Figuratives de M. Minard, 1845-1869, a portfolio of Minard’s statistical maps at the Bibliothéque de I’Ecole Nationale des Ponts et Chaussées, Paris, 62 x 30 cm, or 25X12 in.

This 1869 map is the last sheet in Minard’s

lifetime portfolio. English translation by Dawn Finley, redrawing by Elaine Morse.

ae .

123 B E A U T I F U L E V I D E N C E

P R I N C I P L E S

OF A N A L Y T I C A L

D E S I G N

124

Figurative. Maye of the

s u c c e s s i v e

Losses in men of the conch Oemy in the eR ussiaw campaign

A812 ~1813

° Za

&

e -

A p a l a n

.

Dawn up by MC. Minrcred, Dnopetse Goneral

of Bridges and Reads

in ebivemeut.

wp by p e c g , C p o e a t y e e e e c t

=

Savio, N o v e m b e r 20,1869.

S e m u m b o a of m

e n preseut are. Kepresented by

the widths of the cofored

Boues a t arate

of one willinnetér for

every tew

thousand men; they are

further weitten

acess ~ M

O S C O W

3, S the Zoned.

The red desiquates the men

who entec ints Xussia, the

Wack those who Leave i

t v __

The information which

has cored 6

draw up the map

has been exteactéa f r o m t h e works of

A C M .

Thiers, of eegur;

of c F e z e n d a c , of C

h a m b r a y

and the smpublished

diary of J a c o b , pharmacist of the

Oemy simce

October 28

th.

Iu order to letter judge with

the eye the diminutionof the army,

J have assumed that the troops of Prince

Jerome and of Marshal

Davouslwho had been detached at Minak

a n d Moghiles a

n a Bave_rejoined

around O r c h a

and Vitebsk,

haa a l w a y s marched

with the- a

r e n .

Common leagues of France

(Map of Fezeasac)

Moghilew

G R A P H I C T A B L E

of the

temperature in

degrees of

the [Réaumur

thermometer below

zero. R C T

0 0

32

Rain October 18

The Cossacks

pass the

f r o w n

z :

H e m e s

“10-13. 10

Niemen: at

a gallop.

2 C i f ,

e r e s

—

9. N o v e m b e r 9

- e e

20 -25

-13

= 20. November 28

— 21° November 14

ee —

26° December 7

2 net

A -30

-38 -36

— 30.

December 6

Tp. Lith. Regnier ot Dourdet.

Alutog. par Regnier, 8. Pas. S* Harie

S ¢ C % é Paris.

126 BEAUTIFUL EVIDENCE

Carle Figurative. rs peter sucsiins m boas N Otenie Sangsise dans ta compagie re Russie A812 A813. Deasste pac Me. Mimo, Snspecine. Gondeal res Foutd 1 Canis mie hoy 1869.

Kesonibns Wamenss pists som topiioate paca bargeans tes gues cola i raison: na willinte:poue- Bee lle remeron 20 pla tee tases 210 gones. Le ouge diigue is omnes penton en Rests be -ncic ceme-qyi ea tort Les reasiiuoments sprout decir & dee Le cate omits piste

awa Les & MM. Chiers, dechigur; de-Fezendac; de Chambray ale jomcual inti te Tatoly, poacmacien deb Oizmte- depnis fe 28 Octobce-» ve ic: fie jer Dail Aa Sination 2B cone atop gne a cops Iu Tie Dirome-eu bu Marichal Dosoast-qnioasien itd attache tne Mosh? ee Mbilow eeometejeisets Orscha ea Wit, axaiewtovjencs muaccbi ace Laemate.

_4Moscou %

Polotee,

Laat emer de Fram (2c b So Fatman) ee

en degres da thermomiétre de Réaunur au dessous de réro .

cir

Above, the original French version. This statistical map portrays the successive losses of the French Army in Napoleon’s Russian campaign of 1812. The overall toll, French and Russian, was approximately 700,000 to 1,000,000. Described by E. J. Marey as seeming to defy the pen of the historian by its brutal eloquence,’ Minard’s map exemplifies many 1E. J. Marey, La méthode graphique dans

les sciences expérimentales (Paris, 1878), 73:

“Toujours il arrive a des effets saisissants,

mais nulle part la représentation graphique 7 . de la marche des armées n’atteint ce degré

Pri inciple de Compar' isons de brutale éloquence quisemble défier la plume de l’historien.”

of the fundamental principles of analytical design.

Beginning at the left, on the Polish-Russian border near the Nieman River, the thick tan line shows the size of the Grand Army (422,000 men, drawn from all Europe) as it invaded Russia in June 1812, as well as the army’s path. As the soldiers die, the line narrows; the flow-line indicates the number of remaining soldiers at each position on the map. Also shown are movements of auxiliary troops, who sought to protect the rear and flank of the advancing army.

In September, the 100,000 surviving troops reached Moscow, which was by then sacked, deserted, on fire. The departure from Moscow is depicted by the dark lower line, in turn linked to a temperature scale and dates in the statistical graphic at the bottom of the chart. During the retreat, it was bitterly cold and many soldiers froze and starved. Crossing back over the Nieman River with 10,000 survivors, the army struggled out of Russia. At the end, the map notes: “The Cossacks Les Cosaques passent au galop [Russian horsemen] pass the frozen Niemen at a gallop,” driving out & Niémen gele. the scattered remnants of Napoleon’s Grand Army.

PRINCIPLES OF ANALYTICAL DESIGN 127

And it is at the Nieman River, where the invasion began and ended,

that we see a small but poignant illustration of the First Principle for the

analysis and presentation of data:

Show comparisons, contrasts, differences.

The fundamental analytical act in statistical reasoning is to answer the

question “Compared with what?” Whether we are evaluating changes

over space or time, searching big data bases, adjusting and controlling

for variables, designing experiments, specifying multiple regressions,

or doing just about any kind of evidence-based reasoning, the essential

point is to make intelligent and appropriate comparisons. Thus visual

displays, if they are to assist thinking, should show comparisons.

Minard makes several vivid comparisons, some over many months,

others during the course of a few days. At the war's beginning, the army

crossed the Nieman River with 422,000 soldiers, shown by the number

itself and represented by the width of the tan line. At the end, 10,000

soldiers returned back across the Nieman, as indicated by the number

itself and the thin black line concluding the retreat. Only 1 soldier in 42

survived this futile campaign (the widths of the black /tan lines at the

Nieman River were inaccurately constructed in a ratio of 1 to 28). This

graphic comparison at the Nieman summarizes the slaughter, 6 months

from start to finish.

At the Berezina River, another comparison. During the retreat (west-

ward, to the left), the black line crossing the river abruptly diminishes—

from 50,000 to 28,000 soldiers, as 22,000 died in 2 days. This change

in line weight represents, in a blunt and incomplete way, the terrible

events at the Berezina, described by Philippe-Paul de Ségur (aide de camp

to Napoleon, and a source cited by Minard):

It was the twenty-eighth of November. The Grand Army had had two whole

days and nights in which to effect the crossing, and it should have been too

late for the Russians to do any harm. But chaos reigned in our midst....an

immense confused mass of men, horses, and vehicles besieged the narrow

entrances to the bridges and began to flow over them. Those in front, pushed

by the weight of those behind or halted by the river, were crushed, trampled

on, or forced into the ice-filled water of the Berezina. From the horrible,

formless mob there arose a sort of dull murmur which swelled at times to a

wild clamor mingled with groans and awful imprecations. ... The confusion

was so great that when Napoleon himself wished to cross at two o’clock, it

was necessary to use force to clear a passage for him....Then, as in all extreme

situations, hearts were laid bare, and both infamous and sublime actions were

witnessed. Some there were who, determined to pass at all costs, cut a horrible

way for themselves with their swords. Others opened an even crueler road for

their carriages, driving them pitilessly through the helpless crowd, crushing

men and women, in their odious greed sacrificing their companions in misery

to save their possessions.”

2 Count Philippe-Paul de Ségur, Napoleon’s

Russian Campaign (New York, 1958), 246,

249; translated by J. David Townsend from Campagne de Russie, Mémoires d’un Aide de

Camp de Napoléon (Paris, 1824).

128 BEAUTIFUL EVIDENCE

the jones The ed igus the men who etecinte Rustin, Ue Black. vie be lene ites The

Da eter bette judge with the ge tbe Sominaton af the oom 3 Bave-asswmsd tat the temps ‘ snd Meg ad base rened card Orchard VEER bad alloys mand with the- acy.

3

Figurative Map of the ewazstive Lace in men of the Fecnc® Cteny im the eR ussiaw campaign A812 ~1813. rep by oF. Manned, Inoptoe Goncaak of Beidges ant Reads a-edivmet.

Peer be 1 Dupes Gos of Bees ant Nene Fasis, Novensber 20,1869. Theme of en pro ae epee by Use witha of the clo sone nt ste of oe mili fe ee tow Non mn; tay ae fue weiton ace

sanatcn wbich Las seeed t draw up the wap has Been extends feom—te works of SCM. Thiers, of eMegur; of Fezendac; of Chambray aid the wnpublishea riagy of Jacob; phsemacist of the henuy since. October 28th.

of Beiace Sromeand of Maxsbal Daveust soho bad Been detached at inate

% Moscow %

oe

1

CRAPHIC TABLE of the temperature in degrees of the Réanmur thermameter below zero . The Cossacks pase the frocer ome ata gall. | <r

— 26! Decanter 7 > 30! December 6

2 Hevea 1

Eig par gia Bas Fo BOR Tame

Principle 2: Causality, Mechanism, Structure, Explanation

How did nearly everyone in the Grand Army die? The mapped routes

of invasion /retreat show the location of the bad news but do not explain

what caused the deaths. Yet often the reason that we examine evidence

is to understand causality, mechanism, dynamics, process, or systematic

structure. Scientific research involves causal thinking, for Nature’s laws

are causal laws. Medical analysis— prevention, diagnosis, intervention —

requires causal analysis. Reasoning about reforms and making decisions

also demands causal logic. To produce the desired effects, we need to

know about and govern the causes; thus “policy-thinking is and must

be causality-thinking.”*

Simply collecting data may provoke thoughts about cause and effect:

measurements are inherently comparative, and comparisons promptly

lead to reasoning about various sources of differences and variability.

A vivid statement about epidemiological evidence by the distinguished

medical pathologist Rudolf Virchow exemplifies this logic: “Medical

statistics will be our standard of measurement: we will weigh life for

life and see where the dead lie thicker, among the workers or among

the privileged.”*

Principles of design should attend to the fundamental intellectual

tasks in the analysis of evidence; thus we have the Second Principle for

the analysis and presentation of data:

Show causality, mechanism, explanation, systematic structure.

3 Robert A. Dahl, “Cause and Effect in

the Study of Politics,” in Daniel Lerner,

Cause and Effect (New York, 1965), 88.

4 Rudolf Virchow, Die Medizinische Reform

(Berlin, 1848), no. 1, 182; as quoted in Paul Farmer, Infections and Inequalities:

The Modern Plagues (Berkeley, 1999), 1.

PRINCIPLES OF ANALYTICAL DESIGN 129

Minard depicted a possible causal variable by means of a graph of

temperatures during the retreat—for Napoleon was defeated not only by

the Russian Army but also by General Winter. Temperature is measured

ona Réaumur scale, with the freezing point of water equated to 0° and

boiling point of water to 80°. This is converted to the nearly universal

Celsius (0° and 100°) and to Fahrenheit (32° and 212°) in scales on our

English translation of Minard’s graphic at left. Note that the scale starts

at freezing; the warmest reading is 0°C or 32°F. It was very cold in Russia

during November and December of 1812. The statistical graphic then

brings in an another variable, time. Each temperature during the retreat

is dated, from the rainy October departure from Moscow to the bitter

cold on December 7th near the end.

The map’s causal analysis is thin, merely a reminder of intense cold.

Although the path of the retreating army loosely parallels the line of

falling temperatures, this symbolic parallelism has no evidential value.

The link between very cold temperatures and deaths is, however, made

explicit in Minard’s 5 historical sources that contain eye-witness accounts

of the ghastly frozen soldiers. And in the striking visual-quantitative

imagery of Victor Hugo’s poem L’Expiation (at right) describing the

retreat from Moscow.

Principle 3: Multivariate Analysis

Minard delineates the war of 1812 by means of 6 variables: the size of the

army, its two-dimensional location (latitude and longitude), the direction of

the army’s movement, and temperature on various dates during the retreat

from Moscow. These 6 dimensions are shown with distinct clarity. There

is no instruction manual, nor a jargon fog about a spatial-temporal hyper-

space focus group executive dashboard web-based keystone methodology.

Instead it is War and Peace as told by a visual Tolstoy.

Accounts of Napoleon’s invasion of Russia are multivariate (that is,

involving 3 or more variables). How could it be otherwise? Nearly all

the interesting worlds (physical, biological, imaginary, human) we seek

to understand are inevitably multivariate in nature. Simple navigational

instructions reflect up to 4 dimensions, describing routes through the

3-space in which we live over the 4th dimension, time. Current-day

cosmological theories claim an 11-dimensional universe. The analysis

of cause and effect, initially bivariate, quickly becomes multivariate

through such necessary elaborations as the conditions under which the

causal relation holds, interaction effects, multiple causes, multiple

effects, causal sequences, sources of bias, spurious correlation, sources

of measurement error, competing variables, and whether the alleged

cause is merely a proxy or a marker variable.*

Snow, in what was left of the city behind them,

drifted into the smoke and flames.

In their retreat, the army did not know one white field from another, or the left flank from the right or center. Ensigns whited out, the voices of commanders lost, what had been an army was a herd... .

The sky made turbid snow,

over the largest army ever,

an immeasurable shroud. Each of the soldiers fell

alone, struck by raiding troops,

or by the deadlier North. They junked the cannon to burn the carriage. Whoever slept for a moment died. The wasteland

swallowed them whole, regiments at a time,

visible now, where they lay down to rest,

as undulations in the anonymous snow. Fugitives, wounded men, and dying,

in caissons, stretchers, and sleds, overloading

the bridges, falling asleep by the ten thousand, woke up, hundreds, or less.

Selected Poems by Victor Hugo (New York, 2002), 14-17, translated by Brooks Haxton.

5 On multivariate subtleties of evidence for causality, Austin Bradford Hill, “The

Environment and Disease: Association or

Causation?,” Proceedings of the Royal Society

of Medicine, 58 (1965), 295-300; posted at www.edwardtufte.com

130 BEAUTIFUL EVIDENCE

The only thing that is 2-dimensional about evidence is the physical flatland of paper and computer screen. Flatlandy technologies of display encourage flatlandy thinking. Reasoning about evidence should not be stuck in 2 dimensions, for the world we seek to understand is profoundly multivariate. Strategies of design should make multivariateness routine, nothing out of the ordinary. To think multivariate, show multivariate; the Third Principle for the analysis and presentation of data:

Show multivariate data; that is, show more than 1 or 2 variables.

Principle 4: Integration of Evidence

Minard brings together various modes of information in order to describe troop movements and the war’s consequences: a paragraph of words, a map with narrating flow-lines, and a statistical graphic dangling from the map. In particular, at the Nieman River where the invasion began and ended, these modes of information are simultaneously mobilized: the number of troops before and after the campaign, mapped flow-lines moving in 2-space, temperature and dates, and the mocking words describing the Cossack horsemen. Contrast this graceful integration with books where all the images are collected into one big pile, printed and positioned far from their relevant text. Or those awful reports consisting entirely of words, except for all the data tables segregated and appended at the back, in organizational and intellectual disarray. Or computer applications that deal with only one type of information at a time. The evidence doesn’t care what it is—whether word, number, image. In reasoning about substantive problems, what matters entirely is the evidence, not particular modes of evidence.

To integrate word and diagram, Minard uses a light unsaturated tan color to depict the invasion flow-line, which allows the place-names to show through. The tan’s transparency has little visual effect on the typography for the place-names inside the flow-line, so those names belong with names outside the flowline, maintaining the map’s surface

PRINCIPLES

coherence as the army passes over the land.’ In 20 different figurative

maps, Minard uses transparent overlays to combine words, numbers,

and flow-maps within a common visual field. Cartographers have long

used this technique of layering and separation, which can increase the

informational depth of flatland by additional dimensions and possibly

by several multiples of data density.

Words, numbers, pictures, diagrams, graphics, charts, tables belong

together. Excellent maps, which are the heart and soul of good practices

in analytical graphics, routinely integrate words, numbers, line-art, grids,

measurement scales. Rarely is a distinction among the different modes

of evidence useful for making sound inferences. It is all information after

all. Thus the Fourth Principle for the analysis and presentation of data:

Completely integrate words, numbers, images, diagrams.

Thus tables of data might be thought of as paragraphs of numbers,

tightly integrated with the text for convenience of reading rather than

segregated at the back of a report. Most images and tables used in public

presentations should be annotated with words explaining what is going

on. In exploratory data analysis, however, the integration of evidence

needs to be thought through. Perhaps the numbers or data points may

stand alone for a while, so we can get a clean look at the data, although

techniques of layering and separation may simultaneously allow a clean

look as well as bringing other information into the scene.

More generally, the principle of information integration points to

a philosophy of inquiry: a broad, pluralistic, problem-directed view of

what constitutes the scope of relevant evidence. Too often in scholarly

research, in social science at least, there is a certain narrowness in the

choice and use of evidence. Thus many investigations of, say, political

economy rely exclusively on a single mode of evidence: statistical data,

or wordy memoirs of policy-makers, or anecdotes, or mathematical

models, or metaphor, or economic or political ideology, or newspaper

clippings. Research questions are framed along the lines of “How can

one type of information or one particular approach be used to explain

something?” rather than “How can something be explained?”’

Pre-specifying the mode of relevant information or the explanatory

method may produce a tendentious misalignment of evidence in relation

to substantive matters under investigation. The world to be explained is

indifferent to scholarly specialization by type of evidence, methodology, or

disciplinary field. A deeper understanding of human behavior may well

result from integrating a diversity of evidence, whatever it takes to explain

something. Like good information displays, explanatory investigations, if

they are to be honest and genuine, must seek out and present all relevant

evidence regardless of mode.

OF ANALYTICAL DESIGN 131

© The text accompanying the map describes the invasion line as rouge or red. It appears the line never was red, even when freshly printed in 1869. Although red pigments are degraded by light, Minard’s maps are in fact bound in a non-circulating portfolio with no light exposure at the Bibliothéque de l’Ecole Nationale des Ponts et Chaussées. Today the invasion-line color is as exactly as shown in this chapter (confirmed by color control bars in our direct photograph

of Minard’s original map), with no signs of

color fading. In another map by Minard (below), bright red ink has now survived 148 years, again suggesting that today’s tan

invasion line was not yesterday’s rouge:

Detail from Charles Joseph Minard, Carte ‘figurative et approximative des quantités de viandes de boucherie envoyées sur pied par les

départements et consommeées a Paris, 1858.

7 See Donald P. Green and Ian Shapiro,

Pathologies of Rational Choice Theory (New Haven, 1996). Amusing parodies of a priori method-driven approaches are found in C. Wright Mills, The Sociological

Imagination (New York, 1959) on the social

sciences; Frederick C. Crews, The Pooh

Perplex (New York, 1963) and Postmodern

Pooh (New York, 2001) on literary criticism; The Sokal Hoax, compiled by the editors

of Lingua Franca (Lincoln, Nebraska, 2000) on literary-sociological studies of science.

132 BEAUTIFUL EVIDENCE

Principle 5: Documentation

The credibility of an evidence ptesentation depends significantly on the quality and integrity of the authors and their data sources. Documentation is an essential mechanism of quality control for displays of evidence. Thus authors must be named, sponsors revealed, their interests and agendas unveiled, sources described, scales labeled, details enumerated. Thorough documentation is a good sign that a report was constructed with at least some care and craft.

Minard documents his data-map of Napoleon’s Russian campaign at a level of detail remarkable for 19th-century graphics and in ways appropriate to nearly all visual presentations of information:

What is the display about? Loaseo in men of the tench Cemy in the Russian Campaign A812~A813. Who did the work? Qaawn-up by IM. Mamazd

Who's that? Jnopector- General of Bridges amd Koado inm-_retiremen.

Where and when was the work done? Laws, November 20,1869.

What are the data sources? The: information which has dered to caw up the map bas beow exlzacted from the works of MM. Thiers, of chegur; of Fezendac; of Chambray aid the unpublished diacy of Jacoby toe phanmacist of the Oemy since October 28th.

Any assumptions? Dw oderte better judge with the eye the diminution of the anny, J have assumed that’ the troops of Lrince Jeromeand of Marshal Davoustwho haa been detached at Minsk and Mogbileo and have-rejomed around Orcha aud Vitebsr, haa always marched with the- amy.

What are the scales of measurement?

for invasion and retreat flow-lines: one willimelic for every tew Uousand mew

for the underlying map: Common leagues of France (Map of Fexensac:) The French “lieue” is of variable length, ae ; approximately 2.5 miles or 4 kilometers 6 Se OE Wo. (historically varying by up to 10%).

for the temperature: degrees of the Réaumur thermometer Below zero

Who published and printed the work? Alutog. par Regnier, 8. Pas. St Marie S*C% 3 Paris. Imp. Lith. Regnier et Dourdet.

Publicly attributed authorship indicates to readers that someone is taking responsibility for the analysis; conversely, the absence of names signals an evasion of responsibility. Readers can follow up and communicate with named sources. Also, names may have reputations for credibility—or not.

PRINCIPLES OF ANALYTICAL DESIGN 133

Authorship credit is to often absent from corporate and government

reports; we should remember that people do things, not agencies, bureaus,

departments, divisions. People may do better work when they receive public acknowledgement and take public responsibility for their work. The good Minard put his name on nearly all his work and personally signed with pen and ink (at right) some of the figurative maps.

Displays should name their data sources. Minard names 5 sources.* Information comes from somewhere; the audience should be told where.

Undocumented displays are inherently suspect. For example, opaque, vague, and undocumented accounting statements are leading indicators

of financial corruption.

Data graphics are data graphics because they have scales of measurement. Viewers should be told about measurements. Minard provides 3 scales. Many modern graphics are undocumented: in 13 computer-science

books on technical visualizations, only 20% of the published images had complete scales and labels, 60% had no scales or labeled dimensions at all.” Even in real science, images sometimes forget about measurement

scales: the astronomical pictures from the Hubble Space Telescope

are usually published without indications of distance, size, or location.

In business and financial displays, the common errors and lies involve corrupt measurement scales: absence of labels, undefined or imprecise

measurements, tendentiously chosen base-years, excessively short time-

series, inflated rather than inflation-adjusted monetary units, and time-

shifting of data (such as the notorious premature revenue recognition).

Finally, for many presentations, there are issues of research integrity

provoked by financial sponsorship, advocacy position, and potential conflicts of interest. Students of bureaucracy have long noted that where people stand depends upon where they sit. In medical research, the published findings are too often predictably related to the interests of the financial sponsors of the research. To assess the credibility of a report,

the audience must know who paid for the study, why they paid for it,

and possible conflicts of interest of the authors. High-quality medical

journals now require from authors a detailed statement about financial

sponsorship and possible conflicts of interest, which is then published

along with the research article.'? Public documentation of sponsorship will not end corrupt presentations, but at least it alerts everyone to

possibilities of bias. As J. P. Donleavy said about expecting fair play in high places: “You'll get it if enough folk are watching."

Documentation allows more effective watching, and we have the Fifth Principle for the analysis and presentation of data:

Thoroughly describe the evidence. Provide a detailed title, indicate the authors and sponsors, document the data sources,

show complete measurement scales, point out relevant issues.

repriventéte par des xines maritimes verter entre: des lignes & dont tee largeurs indiguent’ lar tonnages diporis U Echelle.

Paris, le 1¢ Juillet 1855:

Bh

Detail from Carte figurative et approximative de céréales qui ont circulé en 1853 sur les voies d’eau et de fer de l’Empire francais, 1855.

® Minard provides last names; his § sources are very likely Adolphe Thiers, Histoire du Consulat et de Empire (Paris, 1862); Philippe- Paul de Ségur, Campagne de Russie, Mémoires

d'un Aide de Camp de Napoléon (Paris,1824);

M. de Fezensac, Journal de la campagne de Russie en 1812 (Paris, 1849) and Souvenirs militaires (Paris, 1863); Georges, Marquis de Chambray

(anonymously published) Histoire de l’expedition de Russie (Paris, 1823); Pierre-Irénée Jacob’s

long unpublished journal, which eventually appeared in Revue d'Histoire de la Pharmacie,

XVII (1966), numbers 189-191.

° Data in Edward Tufte, Visual Explanations (Cheshire, Connecticut, 1997), 21-25.

10 For example, this statement below from Wayne A. Ray, et al., “cox-2 selective non-

steroidal anti-inflammatory drugs and risk of serious coronary heart disease,” The Lancet,

360 (October 5, 2002), 1073: Contributors W.A. Ray, M. R. Griffin, C. M. Stein, and P. G. Arbogast designed the study and wrote the protocol. W. A. Ray and P.G. Arbogast did statistical analyses.J. R. Daugherty and K. Hall contributed to the protocol and computer programming. W.A. Ray wrote the initial draft of the report, then revised by M. R. Griffin, C. M. Stein, and P. G. Arbogast.

Contlict of interest statement W. A. Ray has worked as a consultant for Merck Research

Laboratories within the past 3 years, but not within the past year. M. R. Griffin is a consultant for Merck. No funding for this study was provided by any pharmaceutical company.

Acknowledgments The work was supported in part by the Agency for Healthcare Research and Quality, Center for Education and Research on Therapeutics cooperative agreement (grant HS1-0384), US Public Health Service grant HL 67964, and a cooperative agreement with the Food and Drug Administration (FD-U- 001641). Study sponsors had no role in study design, data collection and analysis, interpretation, or writing the report.

11], P. Donleavy, The Unexpurgated Code: A Complete Manual of Survival and Manners (New York, 1975), 279.

134 BEAUTIFUL EVIDENCE

Principle 6: Content Counts Most of All

How Charles Joseph Minard (1781-1870) came to construct the map of the French invasion of Russia tells us about the spirit behind one of the best graphics ever. Minard was a renowned engineer, who managed large public works projects and designed bridges, canals, docks, and roads. He had the technical skills to construct good data graphics; he could draw and see and count all at a high level; he had developed the ow-map, an architecture exactly right for depicting losses in military

campaigns. But at the heart of the work was Minard’s passion about the substantive content.

At age 70 Minard retired from engineering. During the next 18 years, he produced 50 maps depicting statistical data describing spatial flows and distributions of French wine, ancient languages, railroads, cotton, migration, and the military strategies of Charlemagne and Napoleon. These “figurative maps” represent Minard’s substantial contribution to the practice and theory of analytical graphics.1?

In November 1869 at age 88, Minard published his last figurative map, the extraordinary 2-color lithograph at right. Two maps are printed on a poster-sized piece of paper. Both maps have the same theme, human losses in war. The upper map recounts the diminishing army of Hannibal and his elephants wandering around the Alps into northern Italy 2,200 years ago. This account of Hannibal’s unhappy adventure makes comparisons over space and time with a narrative flow-line, uses multivariate data, has a north-arrow (documentation that should appear on every map), and indicates data sources. But this story lacks the narrative power of the Russian invasion map.

Minard had seen the horrors of war. Several of his public works projects had repaired wartime destruction, Then, while planning and managing construction in 1813, he was confined in besieged Antwerp:

Minard always retained a sharp impression of some bloody episodes of the bombardment; and these are some memories which had made him leave Paris last year at the approach of the Prussians.

Minard’s last work is an anti-war poster. A memoir by his son-in-law, Minard’s only biographer, describes the map of Napoleon’s invasion:

He emphasized the losses of men which had been caused by two grand captains, Hannibal and Napoleon ist, the one in his expedition across Spain, Gaul, and Italy, the other in the fatal Russian campaign. The armies in their march are represented as currents which, broad initially, become successively thinner. The army of Hannibal shrunk from 96,000 men to 26,000, and our grand army ftom 422,000 combatants to only 10,000. The image is gripping; and, especially today, it inspires bitter reflections on the cost to humanity of the madnesses of conquerors and the merciless thirst for military glory."

? Minard’s work is catalogued in Arthur H. Robinson, “The Thematic Maps of Charles Joseph Minard,” Imago Mundi, 21 (1967), 95-108; Gilles Palsky, Des chiffres et des cartes: Naissance et développment de la car- tographie quantitative francaise au XIXe siécle (Paris, 1996); and Michael Friendly, “The Graphic Works of Charles Joseph Minard,” posted on the internet. Reproductions of Minard’s maps and other graphics are in Arthur H. Robinson, Early Thematic Mapping in the History of Cartography (Chicago, 1982); Edward Tufte, The Visual Display of Quantitative Information (Cheshire, Connecticut, 1983, 2001); and Gilles Palsky, Des chiffres et des cartes... (Paris, 1996).

*3-'V. Chevallier, “Notice nécrologique sur M. Minard, inspecteur général des ponts et chaussées, en retraite,” Annales des ponts et chaussées, 2 (1874), 1-22, quotation at 6-7. Translation by Dawn Finley. A translation of this 22~page biography of Minard at www.edwardtufte.com.

4 V. Chevallier,“Notice nécrologique sur M. Minard. . .,” 17-18.

PRINCIPLES OF ANALYTICAL DESIGN 135

; 5 ; Ligende. Carte. Sigurative es spéees auccossioed ex hommes de Lccewbe. qe Aanibel. conduisi 2” Copague- Deronibcasg Bais = shew d as ‘atl

ea Diakie- en Maversam Aes Cpaules (delon Lalobe- ). po lchngne 20 gus ce anise 3's

Dresiie- par MH. Minard, Onspecterr. Gazal 308 Donte vx CRansthes ew fictraile. Bogut et il sm Rela ilo Kats 208 ia

sguaberte”’ Prix; £29 Novewbo 18by. te SS Ree

sel SPAGNE C4 y Z

: : vr +. fyen Hergetes & pis

. a TRAN SAY Cosy Bargusiens & Fer Bes _

Z 3

Carte Figurative. 20 spies sncsties a bonaes uh! Conde Fangsite rant fae conpagie 2 Russie. A812 “A813.

Peaterpae Me, Minced, Daoptin: Gtk rifts Chava Aste, 95 Noveubre 1869. Seananibcs Vomames psoas Ae papeteis patclsNaegens ts genes elec i ralton- Done williatie pour dscnille roms; its sou— depute m Vanes

sues. Se touge isiguerles hommes quiveutiens on Rustic Le-noitcews-5 wien orto <—__ Ses neusei ion tect, & deesece fa. carte. om ele prise

1 ge Mt Blips de Tizond; dc Chambhay wkjoual iin te Catal, plantas il Cate tpi 28 Oates So iene fae jager a el Aa Dimination IN acne, jsiareppost que es corp Om Fe etn Manica Dawousrqui avaienr ite detaches sue Minsk

_MoscoU

Poloteh: eee

Liner commana de Fre (cts de Xe Kemer) eS

TABLEAU CRAPHIOUE della températute en degrés dy Chermanttre de Réanmar an dessous tie zero.

nt ee ee B

Tes Capees parent 00 gelop b Mince pe. Kit ee 9 his

= wane fs 20" le 28 9 tk 4 9

. aE EXT seat

—t'e7 pe sere x

acy por Regia, b Pas, 5 Mars BOM EPs. Top lal Raper Dear

Charles Joseph Minard, Tableaux Graphiques et Cartes Figuratives de M. Minard, 1845-1869, item 28 ina portfolio of Minard’s statistical maps at the Bibliothéque de I’Ecole Nationale des Ponts et Chaussées, Paris. 62 x $4 cm,

or 25 X 21 in, redrawn.

136 BEAUTIFUL EVIDENCE

ee ins de Sauget i deat La Campaguc ae Russie A8i2 ~A813. Deesiie pac Snpeceu ee tte oy dg 1869. Kesnones Themes pte some apis parts League 0 sue elec i raaon 9 i

Bed Sones. Le sigue Ls Demanes.qieutiont fe-moi. pai em octet Giguementd qui ome—decei & deesoee Le caste oma ite. puisde- Mj lemwye © aM Ch, ler am de Shanty ra Te :

[ Carte Figurative. ses pcs snzstins bonnes do

MOSCoD %

Leet coment th rast Ht Ramee)

| TABLEAU CRAPHAQUE den tempat en dees fiecmanitre de Réaume an dessons de zero. ae a — saan te I val aan Stuer

ere ati i -u7 - oa 30k 6 x*

Pe pe t

| Aap lage bas TP Mane BORER

Ep ga oae

Minard’s focus on the human costs of war is subtly reinforced by his choice of content. Minard never mentions Napoleon. That the word “Napoleon” does not appear on the map of Napoleon’s march indicates here at least full attention is to be given to memorializing the dead soldiers rather than celebrating the surviving celebrity.

Minard’s work exemplifies the spirit behind excellent analytical graphics: a good knowledge of the content and a deep caring about the substance. The Sixth Principle for the analysis and display of data:

Analytical presentations ultimately stand or fall depending on the quality, relevance, and integrity of their content.

This suggests that the most effective way to improve a presentation is to get better content. It also suggests that design devices and gimmicks cannot salvage failed content.

The content principle points to ptiorities in analytical design work: this is a content-driven craft, to be evaluated by its success in assisting thinking about the substance. Thus the first questions in constructing analytical displays are not “How can this presentation use the color purple?” Not “How large must the logotype be?” Not “How can this presentation use the Interactive Virtual Cyberspace Protocol Display Technology?” Not decoration, not production technology. The first question is What are the content-reasoning tasks that this display is supposed to help with? Answering this question will suggest choices for content elements, design architectures, and presentation technologies.

PRINCIPLES OF ANALYTICAL DESIGN

Relevance of Principles of Analytical Design

THE purpose of an evidence presentation is to assist thinking. Thus presentations should be constructed so as to assist with the fundamental intellectual tasks in reasoning about evidence: describing the data,

making multivariate comparisons, understanding causality, integrating a diversity of evidence, and documenting the analysis. Thus the Grand

Principle of analytical design: The principles of analytical design are derived from the principles of analytical thinking. Cognitive tasks are turned into principles of evidence presentation and design.

If the intellectual task is to make comparisons, as it is in nearly all data analysis, then “Show comparisons” is the design principle. If the

intellectual task is to understand causality, then the design principle is to use architectures and data elements that show causality. The Grand Principle helps answer the most difficult question of all in the theory of analytical design: How are principles of design derived?

Tue fundamental principles of analytical design apply broadly, and are indifferent to language or culture or century or the technology of

information display. Nearly everyone everywhere, one way or another, reasons about causality, makes comparisons, navigates through 3-space

and time. The principles are applicable to the design of the first map scratched into stone 6,000 years ago, and also to modern scientific

displays. On both that stone and that computer screen, it is necessary to escape the flatland surface to compare multivariate data, to integrate and

document evidence, to reason about dynamics, mechanism, causality. Perhaps these intellectual tasks and their consequent design principles

have a grander universality: if we to ever see the analytical presentations of intelligent beings from other planetary systems, those designs will

make multivariate causal comparisons. Competence in these fundamental intellectual tasks is what it takes for sentient and communicating beings to survive, thrive, and evolve in a world governed by the universal laws

of Nature — for the character of Nature’s laws is that they are causal,

operate at every multivariate space-time point, reveal themselves by

difference and comparison, and, indeed, are utterly indifferent to what

particular sentient beings think about those laws. The universal character

of physical laws provokes certain universal analytical tasks, here turned into the principles for the presentation of evidence.

Because these principles are rooted in fundamental cognitive tasks, they are relevant for producing presentations and for consuming presentations. Thus consumers of evidence presentations should look for appropriate comparisons, assessments of causality, multivariateness, use of relevant

data, credible documentation, content-reasoning. There is a symmetry of thinking in the wise production and the wise consumption of evidence.

At a good evidence presentation, we’re all in it together.

137

138 BEAUTIFUL EVIDENCE

What about analytical displays of evidence involving human behavior, often so distant from any kind of lawlike understanding? Social science lacks the wonderfully convenient simplifying guarantee of the physical sciences that empirical observations are always somehow the expression of knowable invariant laws that operate at every space-time point. Lacking this mind-focusing guarantee, the study of human behavior is sometimes overwhelmed with multivariate uncertainties about causality, as analysis of after-the-fact historical patterns yields loose, fragile, poorly resolved explanations. Consider medical research seeking to identify successful treatments, a field advantaged by a wealth of resources and tewards, by clear explanatory goals, by the capacity to conduct random- ized controlled experiments, and by a basis in universal physical laws. Even with such advantages, the substantive results in medical research tend toward the marginal and unsatisfying (“21% of the treatment group survived compared to 15% of the controls”) rather than the certain (“this treatment always produces a cure”).

Indeed, perhaps the single most important finding of social science— the unanticipated consequences of human action—is a finding about the impediments in predicting consequences of human behavior.'® Such uncertainty is epidemic because people can act on the basis of knowledge about patterns of human behavior, and thereby modify (uncertainly, to be sure) those patterns. Many investigations of human activities seek to unveil particular historical experiences in order to alter, avoid, or continue those experiences in the future. Thus the hope of Minard in making his anti-war poster.

Compared to evidence presentations about nature (physical science), presentations about human behavior (medicine and social science) are more descriptive, more verbal, less visual, less quantitative. For example, the physical and biological sciences publish statistical graphics that depict a median > 1,000 numbers (data for Nature and Science, 2006). In contrast, for applied research in medicine, statistical displays show an average 45 numbers per graphic (data for The Lancet, 2006). Also evidence published in major scientific journals such as Nature and Science is distinctly more visual and nonverbal than the evidence published in major journals in medicine and social science.

Despite such differences in modes of evidence and in explanatory style, physical science and studies of human behavior share some key underlying explanatory and intellectual patterns. To this point, the editor- compiler of 1,045 grand summary findings about human behavior that had some reasonable supporting evidence offered a threefold grander summary of social science knowledge:

(1) Some do, some don’t.

(2) The differences aren’t very great.

(3) It’s more complicated than that.**

15 Robert K. Merton, “The Unanticipated Consequences of Social Action,” American Sociological Review, 1 (December 1936),

894-904.

16 Bernard Berelson, “Citation Classic: Bernard Berelson and Gary Steiner, Human

Behavior: An Inventory of Scientific Findings (New York, 1964),” in Neil J. Smelser, ed., Contemporary Classics in the Social and Behavioral Sciences (Philadelphia, 1987), 335.

PRINCIPLES OF ANALYTICAL DESIGN

These summary statements reflect the universal analytic issues of

(4) causality, (2) comparison, and (3) multivariate complexity. Human

activities, after all, take place in intensely comparative and multivariate

contexts filled with causal ideas: intervention, purpose, responsibility,

consequence, explanation, intention, action, prevention, diagnosis,

strategy, decision, influence, planning. Thus the fundamental principles

of analytical design are relevant for displays of evidence describing

human behavior, as we have seen in Minard’s French Invasion of Russia.

139