Status Updates From The Visual Display of Quant...

In a "friendly", accessible graphic, labels are placed on the graphic itself, elaborately encoded colors are avoided, and *no legend is required*.

Tufte has a gorgeous example of text-graphic integration in Leonardo da Vinci's notebooks.

Tufte suggests that the segregation of text and tables/graphics is mostly an accident of the development of printing technology (and now, I guess, universal plotting software like matplotlib/plotly/R). I can see that integration would be good: imagine plots labeled in beautiful LaTeX in ML papers instead of pyplot's gross sans serif.

The same typeface should be used for graphics and text.

Tufte doesn't like references to figures like "See Fig 1". As much as possible, tables and graphics should run into the text. Morever, "[i]f a display is discussed in various parts of the text, it might well be printed afresh near each reference to it, perhaps in reduced size in later showings".

This is fairly unconventional advice but sounds solid.

Tufte prefers a "supertable" to "a hundred little bar charts". For what it's worth, I disagree. In a bar chart, you can visually discern how big a datapoint is *relative* to another. You can quickly notice a datapoint that's much larger, for example. Reading visuals is faster than reading numbers.

Tufte comes down strongly against pie charts:

> A table is nearly always better than a dumb pie chart; the only worse design than a pie chart is several of them, for then the viewer is asked to compare quantities located in spatial disarray both within and between pies ... Given their low data-density and failure to order numbers along a visual dimension, pies should never be used.

He even cites a 1981 Bertin book.

By simply throwing more data at you, sparklines help "find an approximate answer to the right question (rather than an exact answer to the wrong question)" (since you don't have to bet on the right summary statistic).

Incorporating sparklines into tables of financial data seems like a great idea. First, patterns are jusr easier to see visually than lexically. Second, tables can only show current changes, potentially causing recency bias. Sparklines "improve the attention span of tables".

The idea of sparklines as "datawords" is quite cool. They also look incredibly well-designed. The color choice connecting the last datapoint to the printed number, and the deliberate compactness, are especially pleasing.

Along with the "lie factor" and "data-ink maximization", the idea of high-resolution data seems like one of the most notable in the book.

Except Minard's chart of Napoleon's conquest that everyone and their mom fetishizes, of course.

Tufte prefers that graphics maximize data per area. This can be done in two ways. First, expand rhe "data matrix", ie, the matrix of variables and datapoints. Instead of summary statistics, show real datapoints, IQR, and min-max. Second, shrink the chart's size. This actually makes patterns easier to notice in the examples Tufte provides.

> The simple things belong in tables or in the text; graphics can give a sense of large and complex data sets that cannot be managed in any other way.

I got a sort of Tetris Effect from reading this book, where graphics I saw in real life became much more noticeable. Especially their data-"thinness".

this is legitimately a really interesting book

19 pages read in 8 min, via @leioapp (http://leio.co).

12 pages read in 5 min, via @leioapp (http://leio.co).

Recommended by my favourite professor and gifted to me by my mom, I'm very excited to finally read this and learn more about the magic of great data graphics.

Extraordinary

I really need to make a “This book says John Snow ended the cholera epidemic by removing the pump handle but he didn’t” shelf.

Probably, indeed, the larger part of the labour of sifting, combing, constructing, expunging, correcting, testing: this frightful toil is as much critical as creative.

These visuals make your girl happy

Chart junk

Context is essential for graphical integrity

Distortion of data

The principle of graphical excellence.

Narrative graphics on space and time

Time series

Data maps

Still early in the book, first impressions: Tufte puts together a history of data visualization, from its birth in ancient cartography to abhorrent examples of statistical distortion illustrating newspaper articles. Here's hoping I'll learn some useful things for developing my UX and dataviz skills. Visualization is more than seeing, it's about reasoning based on information that was previously only numbers on tables

Love the joke on page 59. Top notch material.