This graph garnered a lot of opinions about a year ago, setting data vis twitter and blogosphere ablaze. As an attention getter, it certainly was very effective.
In R, there is the spiralize
package that makes making this kind of a visualization very easy. Other platforms may have to work harder to make something like it.
Spiralize might be useful for Du Bois style spiraling bar charts as well.
Jacques Bertin breaks down a graphic display and its design:
On the abstract end, the invariant of the graphic - the idea, concept or topic that unifies all visual marks.
More concretely, the variable features are the components of the invariant.
Each component consists of atomic parts, called elements.
Jacques Bertin breaks down a graphic display and its design:
Two critical ingredients to an element:
Jacques Bertin breaks down a graphic display and its design:
Six fundamental visual variables:
Jacques Bertin breaks down a graphic display and its design:
Three stages of reading a graphic:
Jacques Bertin breaks down a graphic display and its design:
Three questions a graphic should answer:
A graphic that can do all three at once is called efficient.
Building further from Bertin’s abstraction, the community has developed a richer language of visual marks and their properties.
Munzner discusses points, lines and areas.
Wilkinson discusses a wider range, including intervals, paths, schemas; polygons, contours; and edges - all of them in 1, 2 or 3 data dimensions.
Next week we will look deeper into various properties of geometric marks, and the perceptual efficiency of communicating different kinds of information through the different visual channels.
Each of the 4 levels suggests different failure modes, different modes of study, analysis and validation, and different approaches to improving a design.
Pick one of the following papers and skim it to find their description of how they validate their designs. Which of Munzner’s levels do they work at?
All papers available online through CUNY’s library systems.
Skim the paper, and extract their validation methods.
Discuss with everyone who have read the same paper, check that you have compatible understandings.
Form a group with only people who have read the other two papers, compare validation designs.