The Visual Display of Quantitative Information

danny writes "The Visual Display of Quantitative Information is a recognised classic on statistical graphics; to the 1983 original, this 2001 edition adds some additional graphics, extra colour, and corrections. It is a broad-ranging work, covering history, theory and practice and, despite the formal title and scholarly references, not at all narrowly academic. It assumes only a very basic understanding of statistics." Read on for the rest of Danny's review. The Visual Display of Quantitative Information author Edward R. Tufte pages 197 publisher Graphics Press 2001 rating 10 reviewer Danny Yee ISBN 0961392142 summary the classic work on statistical graphics

Tufte begins with the different kinds of informational graphics (maps, time-series, narratives, and relational graphics), describing their origins and evolution and presenting examples of excellence in their design. Many of these are fascinating in their own right -- two that I particularly appreciated were Minard's depiction of Napoleon's disastrous retreat from Moscow and an 11th century map of China.

"For many people the first word that comes to mind when they think about statistical charts is 'lie.'" Tufte gives examples of different kinds of deceit in graphics, along with some principles for maintaining graphical integrity. He goes on to consider the reasons for the poor quality of many informational graphics: one is the relegation of their design to those with art training but without an understanding of either the substance of the material or of quantitative (statistical) methods.

Part two begins by introducing some terminology and theory for describing graphics. The principle "Above all else show the data" is formalised as maximization of the data-ink ratio, and illustrated with some "before and after" examples of erasure of redundant or non-data-ink. Tufte excoriates various kinds of "chartjunk": moire vibration (the disconcerting effect caused by repeating patterns), the overuse of grids, and the "ducks" created when the design takes precedence over everything else.

Tufte gives specific suggestions for the design of box plots, bar charts, and scattergraphs. He argues for the use of multifunctioning graphical elements -- building data measures or grids out of the data itself, for example, by using labels that also show the end points of the data ranges. And he looks at ways of maximizing data density (within reason) and using "small multiples," or repeated smaller graphics. A final chapter steps back to consider the balance between text, text-tables, tables, semi-graphics, and graphics -- "Given their low data-density and failure to order numbers along a visual dimension, pie charts should never be used" -- and to touch on the aesthetics of proportion and scale.

All of this is liberally illustrated with examples, drawn from across the natural and social sciences. Despite the space devoted to these, The Visual Display of Quantitative Information packs a lot in, avoiding repetition or verbosity. Tufte's own tables and graphs are appropriately effective and the volume as a whole is elegantly put together: though it's more than that, it could be appreciated simply as a work of art. Tufte also finds room to survey publication practices across a select sample of international newspapers and journals, comparing the data density of graphics and the proportion of relational graphics (involving at least two variables that aren't temporal or spatial).

Most obviously, The Visual Display of Quantitative Information should be read by those involved in writing, editing, or designing documents or displays that contain statistical graphics -- from professional editors, technical writers, academics, and journalists right down to high school students. But others may appreciate it too: it has changed the way I look at informational graphics.

Danny has written over 700 book reviews. You can purchase The Visual Display of Quantitative Information from bn.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the (recently updated) book review guidelines, then visit the submission page.

Comments

Details about what's different since 1st edition?

[dbrower]

I'd like to have some more data about the new information - what has changed, and why. Anybody know?

-dB

Good Book!

[SEGV]

I've mostly read this and Tufte's two other books, which a colleague kindly brought into work. They are all three a joy to read or just look at.

Re:Good Book!

[Stephen]

The other two books are called Envisioning Information and Visual Explanations. I recommend all three of them, although there is some overlap.

Have a look at Tufte's sculptures too.

Re:Good Book!

[Walter Wart]

I would have to add a couple comments here. "Good Book" is a good start, but "Why this is a good book" is useful, too.

Consider a possibly significant piece of information. I am fortunate enough to live in Portland, Oregon, the home of Powell's bookstore. The technical bookstore alone is a cavern covering most of a city block. The main store is three stories tall and does cover a whole block. There are always lots of new copies of Tufte's stuff on the shelves. I have almost never seen a used copy. People buy his work and hang on to it. Anyhow....

TVDoQI, Visual Explanations, and Enivisioning Information are a lot more than how to draw useful graphs. They explain, in some detail, how to communicate potentially complex information succinctly, elegantly and understandably. They also provide excellent heuristics for detecting bad graphical representations and ways of determining when someone is trying to lie to you with pictures. The only books which come close are Schopenhauer's The Art of Controversy for argument and How to Lie With Statistics for statistical skullduggery.

Tufte uses many different styles and examples and arranges them in ways which expand the reader's view of what can be shown as well as how to do it.

I just got a copy of The Cognitive Style of Powerpoint and am very impressed. He lays out very clearly its inherent and possibly insoluble problems vis a vis information density, the abomination that is the content wizard and the colorful graphing tools which hide more than they show. He's on slightly less solid ground with the cognitive style and "straitjacket" comments; they are generally true but not necessarily so.

He's got a website where you can buy his books, monographs, posters, graph paper, and even a quarter million dollar sculpture.

Take One Of Tufte's Courses

[DoctorMabuse]

As a consultant, I have to be very careful when choosing seminars/courses to attend. One, because they cost money, and two, because I don't get to bill that day(s), which even costs me more money.

I will never regret attending Tufte's course. I learned more about web design, the evils of Power Point (see his article in a recent Wired) and other topics related to display of information, than I ever imagined possible. His course isn't for academics. If you ever give briefings where you have to display pie charts or bar graphs, you could learn things from his course. Highly recommended.

Re:Take One Of Tufte's Courses

[CGP314]

The problem isn't with power point, the problem is that is lowers the barrier for an incompetent speaker to put together a speech - a good speaker, however can use power point to do amazing things. IE, just yesterday I was at an hour long presentation where a librarian told us how to use the library. It's was just slide after slide of URLs and paragraphs she read from the screen. On the other hand, a few years ago, I saw an excellent speaker use power point to give a talk on the basics of string theory. He used power point well - as a way of organizing his diagrams and animations.

Re:Take One Of Tufte's Courses

[donpardo]

Tufte makes this point in his presentations. He thinks it's fine for use as a slide projector, but terrible for building the images.

The classic example: Gettysburg

A Vey Useful Book

[TechyImmigrant]

I've used this book for years as my first point of reference when I have 'quantitive information' to display.

Every engineer should have one to hand to keep themselves safe from the brain warping effects of powerpoint.

Re:A Vey Useful Book

[Enonu]

What do you mean?!?!? Just like, uh, get a few sentances, put bullets next to them and *BAM*, you have a presentation for everybody. Just add some swirlin graphics and some sound, and the information really stands out. To boot, since you can spend less than a minute per slide, make like a 100 of them, and then, here's the kicker, make your audience print the whole damn thing! That way they don't have to take notes or think even.

Re:Details about what's different since 1st editio

[zontroll]

info from the author and publisher on what's different since the 1st edition is here.

The Columbia Team could have used this book

[twiddlingbits]

The author was interviewed by the CAIB. He stated a lot of the information presented that lead the NASA managers to the "we can't do anything" approach was poorly designed and did NOT get information across, or was slanted. He pointed out several PowerPoint slides that had 10-12 errors on them that led to incorrect interpretations by the audience. This is a GREAT book, and should be on every engineer's shelf if they present data to an audience (including peers). MBA's study some of this in their classes, but (most) Engineers and Scientists and Doctors don't. It's a shame when you have great information that is hidden by poor presentation.

Bad statistical graphics are everywhere

[rev063]

Despite this classic book being available for 20 years, there are still plenty of examples of bad statistical graphics to be found in the mainstream media. Here's just one example from the Seattle Times this year, along with a "non-lying" revision of the chart, but you can find plenty by flipping through just about any regional newspaper. Or any edition of USA Today. The NYT usually has good charts, though.

I thoroughly recommend this book to anyone who has to produce charts for their job.

Re:Bad statistical graphics are everywhere

[B'Trey]

The "corrected" chart is worse than the original. The author says "I re-did the chart in Excel, and you can see that a real chart tells a very different picture: actually, it's the middle classes that carry the greatest share of the tax burden, however you look at it."

Balderdash and poppycock. Why are you splitting the top 5% into two groups? Why, to make their contributions look smaller! That's a deliberate deception. How about if you combined the top 2-5% with the top 1% and show the top 5% as one group? The top 5% pays 56.1% of all income taxes. The middle class and the lower class combined pay 43.9% So tell me again who pays "...the greates share?" The rant is titled "Damn Lies." It doesn't bother to inform you that it's the ranter who's telling them.

(And for what it's worth, a bar chart was a poor choice for this graph. A pie chart would be preferred.)

Re:Bad statistical graphics are everywhere

[extra88]

Why are you splitting the top 5% into two groups? Why, to make their contributions look smaller!

Because that's how the original Seattle Times graphs broke them down. And given how large a part of the federal budget is devoted to Social Security and Medicare, it's ridiculous to look only at income tax and not at all federal taxes. The total for all federal taxes for the top 5% is 41.2% of all taxes paid vs. the 44.6% paid by the "middle class" WHICH WERE ENTIRELY OMITTED from the Seattle Times graph.

(And for what it's worth, a bar chart was a poor choice for this graph. A pie chart would be preferred.)

While the fact that they're taking 100% of something and splitting it up implies a pie graph would be favorable, it's easier for people to see differences when it's a bar graph vs. a pie graph.

Re:Bad statistical graphics are everywhere

[B'Trey]

Because that's how the original article (sadly no longer available) divided society.

Funny, I have it up in another tab right now. I followed your link to the Seattle Times, created a bogus login, and accessed the article in the archives. It does not have the graphic in the archives article, but your page links to that.

Nowhere in the article does it divide society up as you claim. I only find one mention of different percentiles:

Seattle Times Quote: The top 5 percent of the nation's taxpayers paid 41 percent of all federal taxes, according to the Joint Committee on Taxation. But that same group paid from 56 to 59 percent of all income taxes.

There were four groups: lower classes (bottom 60% quantiles) middle classes (not mentioned), upper classes (above 5%) and super-rich (above 1%). Then a graph compared income taxes paid by each. The top two groups were compared IN THE CONTEXT OF the bottom group, but the top two groups overlapped each other! This is deception.

This is simply false. The article never mentions three or four groups. I posted the only mention of percentages in a quote above. Pull it up and look. (If you dispute that, let me know and I'll paste the entire text of the article in a comment and copyright laws be damned.) The point of the graphic is not to compare the bottom 60% to the top 5% to the top 1%, nor is it to show the distribution of the tax burden across the entire spectrum of the population. The point of the graph is to compare each of the top 5% and top 1% with the bottom 60%. The fact that the top 5% and top 1% overlap is intentional. No attempt is made to have the numbers add up to 100%. THAT ISN'T THE POINT OF THE GRAPH. In effect, the graph says "The top 5% pays much more that the bottom 60%, and even if you restrict it to the top 1% they still pay more!" The point is to show the disparity, and the graph does that admirably.

Bertin's point regarding multiple pie charts is correct. Also, pie charts with more than four wedges are difficult to comprehend. That doesn't mean pie charts are useless. One would have worked fine for your "corrected" graph. (It would have been a poor choice for the original graph, however, because the original graph was not intended to show a division totaling 100%.)

But it doesn't make sense to compare the ratios of A+B with C, with the ratio of B with C. That's plain misleading.

It makes perfect sense to compare the top 5% with the bottom 60%, then to compare the top 1% with the bottom 60%. That's the purpose of the graph.

And, for the record, IAAS (I am a statistician).

God but I hope this is a lie. If not, please let me know where you earned your degree so I can make sure my kids don't go there.

Tufte's website

[rev063]

Check out www.edwardtufte.com for more information about Tufte's work. The story of how he had to publish his own books to get them looking exactly how he wanted (important when your thesis is about ideal visual display!) is very interesting. The "Ask E.T." section is also well worth a read.

Filelight

[Makarakalax]

I've written a small program for KDE that exhibits what I feel is a fairly novel method for representating hierarchical data graphically.

Currently it only shows information related to your filesystem, but with the next version it will begin accepting any kind of hierarchical data piped from the cli, via a text file, etc. (method of input as yet unfinalised).

If anyone's interested, here's a screenshot, and here's the homepage

I apologise for the plug; usually I'm quite good and wait for at least on-topic opportunities! I'm sure I'll still get the usual ac death threats etc. nothertheless I hope to have interested some people.

Coffee Table Book for PowerPoint Jockies

[Anonymous Coward]

I got this book for Xmas a few years back and was a bit disappointed -- it is basically an "old school" version of Jakob Nielsen. The book was pretty and a somewhat interesting read with (as everyone always mentions) good historical examples, but expensive and ultimately not incredibly insightful. ET seems to have carved a good niche for himself making PowerPoint jockies feel part of a broader cultural tradition.

Sorry to be so negative.

Napoleon map and ClickTracks

[Stephen]

Interesting that you should single out the map of Napoleon's retreat from Moscow. That graphic was the inspiration for our web log analysis program ClickTracks. Our CEO saw it and realised that what web log analysis needed was to show data in context, rather than in long lists. We have the poster of the Napoleon map on the wall of our office.

Link to Article

[Hell O'World]

Read the Tufte article here

Information Visualization Resources on the Web

[Adam_Trask]

FYI: some more resources

Great Theories-- but not for Everyone

[Speare]

I love Tufte's books-- have three of them on my shelf. They're very intricate and excellent. They're visceral studies in how you can achieve excellence.

But that's just the problem: they are over the top. Not everyone is an aesthete.

Most people would rather just type a few lines into a PowerPoint template and flash it onto an 800x600 screen, rather than hire a team of graphic artists to develop a diecut 1200dpi offset-print folder of reports which draw a visual metaphor between daVinci's visions and last quarter's sales in the Kansas region. And most of the time, audiences would rather skim than study, too.

I see Tufte as belonging somewhere between Knuth and Escher. If you consider his valid points and enjoy the energy he brings to the craft, great. But keep pragmatism in the process too: if it's good enough, ship it, and refine it for the next revision.

Re:Great Theories-- but not for Everyone

[dmeranda]

Good presentation of information IS being pragmatic. Too often people substitute the word "pragmatic" for the word "lazy".

Speaking of the PowerPoint generation, as a software developer who actually tries to study information presentation from the likes of Edward Tufte, Jakob Nielsen, and so forth I still get real frustrated when the PHB's dictate requirements with no insight at all. Often times colors choices are made just by picking the prettiest color amongst the 32-color palette available in the MS Word toolbar or something silly like that. I find that in the real world it's not that Tufte is not very pragmatic, its just that the people making the decisions have not even heard of Tufte.

Re:Great Theories-- but not for Everyone

[Linux Ate My Dog!]

Actually, you touch on what I consider a problem with Tufte and the computer domain. Many people always point to Tufte as the examples to follow into making clear displays and websites to display lost of data. But when you actually try to adapt anything he does, one quickly find out that most of his examples of visualizing repetitive data are predicating on using high-resoltuion output -- like paper. Not screen. You cannot show 40 T-shirts, or faces, or blocks depciting small differences in a dataset on a 72 or even 110 dpi screen, at the same time. Unless your screen is a poster. Just can't be done.

In fact, when actually specifically starts to give recommendations working on the screen, they are not so good, although I cannot find the link to such a recommendation that supports my opinion.

Pity about the title.

[rumblin'rabbit]

This book, and its predecessor "How to Lie with Statistics", have been described as "The Elements of Style" for graphics. But William Strunk Jr. would have been horrified by the title. Visual Display? What other kind of display is there? A better title would have been "Displaying Quantitative Information". Having said that, I greatly enjoyed the book, and learned much from the 1st edition. Anyone working with graphics should read it.

Ideals versus reality

[deuterium]

This is another case of people espousing somewhat academic ideals because they're either idealistic or posturing. Much like the methods academics would suggest for general programming, a lot of the examples in this book assume a populace that has the ability to understand them. This isn't to say that we shouldn't know the principles behind information display, but those can be had in much simpler books, with more real world examples. Sure, it's cool that the Okinawa train schedule packs a ton of info into an efficient area, but it's also a daunting read for the uninitiated. People deal best with once piece of information at a time. It reminds me of web sites that use menus based on a 3d sphere of text, or interrelated sliders. It may make perfect sense to some people, but not to all. Truly, KISS.

Re:Worth the read

[HarveyBirdman]

Overheard recently: "I go to customer sites. They show me their Powerpoint presentation. I show them my Powerpoint presentation. And we think we've communicated."

Well that's the fault of "they" and "we", not Powerpoint. People blaming the tool, again.

Do the changes make it worth buying?

[Phronesis]

I don't get any sense from the review whether the changes make it worth buying the new edition if I already own (and have committed to memory) the first edition.

Warning

[rnd()]

There was a story on this book in Slashdot a few years ago, and I ended up deciding to purchase the book.

It's decent material, and it's all accurate, but it's nothing revolutionary.

Generally, when the media publishes a misleading chart or graph it's done intentionally, and anyone of moderate intelligence realizes that when viewing the chart or graph.

The book is like Strunk and White for people who display quantitative information.

Strunk and White is not useful for most people raised on standard English grammar, and is quite frankly annoyingly parochial. Tufte's books strike me similarly. For instance, just as Strunk and White would likely find authors like Jack Kerouak or Junot Diaz abhorrent, Tufte would find Wired magazine abhorrent for all its visual excess and non-information-conveying design.

I haven't seen the latest edition, but I recommend browsing through this one at the book store before spending money on it.