Slopegraphs for comparing gradients: Slopegraph theory and practice

June 1, 2011  |  Edward Tufte
16 Comment(s)

Slopegraphs compare changes usually over time for a list of nouns located on an ordinal or interval scale.

Many examples, the first from my The Visual Display of Quantitative Information (1983):

 image1

This table-graphic above organizes data for viewing in several directions. When read vertically, the chart ranks 15 countries by government tax collections in 1970 and again in 1979, with the names spaced in proportion to the percentages. Across the columns, the paired comparisons show how the numbers changed over the years. The slopes are also compared by reading down the collection of lines, and lines of unusual slope stand out from the overall upward pattern. The information shown is both integrated and separated: integrated through its connected content, separated in that the eye follows several different and uncluttered paths in looking over the data:

 image2

Such an analysis of the viewing architecture of a graphic will help in creating and evaluating designs that organize complex information hierarchically.

Source: Edward Tufte, The Visual Display of Quantitative Information, 158-159.

Below, a table showing cancer survival rates for 5, 10, 15, and 20 years.

 image3

Below, this table-graphic, a slope graph, gives a rough visual idea of time-gradients for survival for each cancer. Like the original table, every visual element in the graphic shows data.

 image4

The slopes could also be sparklines showing how the detailed changes add up to the overall change.

Source: Edward Tufte, Beautiful Evidence, 174, 176.

Bumps charts can also be considered slopegraphs. Here is a spectacular bumps chart from my Beautiful Evidence (2006) pages 56-57.

 image5

The 1987 bumps chart in my Envisioning Information (1990), page 111, also qualifies as a slopegraph.

Although it would be better to include at the right a shuffled list of nouns.

 image6

Can our Kindly Contributors provide other examples?
An obvious candidate is the slopegraph of the average life span in 1980 and in 2010 for 50 or so countries.

January 13, 2014
Slopegraph contest:
Two winners of slopegraph contest, each receives $600 + a set of 4 books

Pascal Schetelat
https://github.com/pascal-schetelat/Slope

Excellent design and interface, good documentation of program, use of color

Ben Concutere

https://github.com/concutere/sg

http://concutere.com/sg/sg.html

Excellent design, good documentation, colors, SVG

Thanks you all! ET and the small commitee of happy judges.
The two winners should get in touch with me at etscuplture ]a t} goomail etc write a program posted on github or as a Chrome app whose output closely replicates the (1) GNP slopegraph and the (2) cancer survival slopegraph immediately below. Necessary subtleties include thin gray lines that don’t crash into words/numbers, typeface Gill Sans or equally refined, tuned ordinal spacing of words at left with the line paths, probably best to make each line graph separately and then order and stack them appropriately to avoid too many line crashes. The idea is to compare slopes, with what are in effect a set of separate plots then ordered by first entry and then stacked with some optical care. User option: assign various clear but quiet colors to occasionally single out a few particular lines of interest. There should be a separate data-documentation box describing the source of the data, a link to the original data set, and the person directly responsible for the data and graph displayed. Much better if your code doesn’t require a lot of extra stuff/apps to run; best if it runs in modern browsers Chrome, Safari, whatever, or as a Chrome app.

You might notify us of your entry by going down to the bottom of this thread to the “contribute” and provide a link to your entry. We will also search the web for good slopegraph programs All this is not easy. But a successful easy-to-use implementation would be a great contribution to open-source statistical graphics. Slopegraphs focus on slopes, deltas, changes; have lots of data; and have a straightforward reading even to the statistically innocent.

$1200 in prizes arbitrarily awarded. thanks, et

Topics: E.T., Science