Brewing Better Charts and Maps 28
Frosty P writes: Cindy Brewer chairs the geography program at Penn State, and has devoted much of her career to helping other people make better maps. By bringing research on visual perception to bear on design, she says cartographers can make maps that are more effective and more intuitive to understand. Many of the same lessons apply equally well to other types of data visualization. A big reason people run into trouble with their color schemes, Brewer says, is the way color picking is done in many software programs. Take the RGB cube (or sliders) many programs use to display colors along red, green, and blue axes, for example. 'That's not the least bit perceptually scaled,' Brewer said. 'In some parts of the cube a tiny step gives you a huge perceptual difference. In other parts it all looks the same.' Brewer's best-known invention is a website called Color Brewer, which helps mapmakers pick a color scheme that's well-suited for communicating the particular type of data they're mapping.
Slashvertising? (Score:2)
Looks like an ad for a technology that's pretty obvious.
Re: (Score:2)
Re:Slashvertising? (Score:5, Insightful)
It's so obvious that I encounter the problem in maps, pie charts, categories in video games, etc.
I guess it's not that obvious after all. Add the various degrees of color-blindness that a lot of people have and you end up with a serious issue.
Re: (Score:2)
Re: (Score:2)
Nope, as a map nut I just thought it was an interesting story with interesting ideas to people who work with maps, UIs, those sorts of things.
Re: (Score:2)
RColorBrewer (Score:3, Informative)
Re: (Score:2)
Ed Tufte (Score:5, Insightful)
A big reason people run into trouble with their color schemes, Brewer says, is the way color picking is done in many software programs. Take the RGB cube (or sliders) many programs use to display colors along red, green, and blue axes, for example. 'That's not the least bit perceptually scaled,' Brewer said. 'In some parts of the cube a tiny step gives you a huge perceptual difference. In other parts it all looks the same.'
I remember Ed Tufte making the same point about displaying information with rainbow color-maps, an unfortunate favourite of many scientists.
Re: (Score:2, Interesting)
an unfortunate favourite of many scientists.
In my experience, it is more likely whatever comes as default with the plotting program they use. After 1000 other things have been painstakingly dealt with, there have been zero papers rejected because they didn't choose the right colormap, so it is inefficient to waste time on it.
Re: (Score:2)
I was involved in research on effective color mapping at Lawrence Livermore 40 years ago. The issues discussed here were ones we analyzed and designed mapping tools to provide mappings that worked well for a variety of applications. Mapping proved closely linked to both the nature of the data (e.g. smooth vs. high contrast edges vs. noisy) and what the goal of the mapping was. I know of photography and graphics designers who still use the papers generated from this research, but mostly it seems to have been
"Software" programs (Score:4, Funny)
Every time I see an article mentioning "software programs" I cringe. I guess they're different from "hardware programs" or "exercise programs" or you can just call it "software" like everyone else. It reads like someone still uses a typewriter and not one of these fancy new computers using "word processing software programs". *sigh*
Re: (Score:2)
'Software programs' is the plural of 'software program'. The word 'software' alone in that context wouldn't work.
If the article just said 'programs', then it could refer to programs in the sense of 'an agreed form [for presenting data]': so it could be about chart presentation in general rather than software design. It would be slightly archaic sounding, but it would be a valid interpretation.
The cringer (diodeus) is perhaps not familiar with such things as concert programs, which are often full-color booklets, or television programs, which have been in color since back in the 1960s or so. In both of these, the colors were originally and sometimes still are rather crappy and unrealistic, so they're on-topic in the current discussion.
Dunno why you'd call them "archaic sounding", though. I've seen lots of concert programs and a few television programs in recent years. The TV is starting to l
Pattern, repetition, variation, and regulation (Score:4, Interesting)
I was struck by the similarities in the shades of colors in the Color Brewer to the use of patterns, repeats, and variations in music. When you hear the same musical pattern repeated over and over, it sets up an expectation in the mind for that pattern to continue. In the Color Blender, choosing a range of mono-chromatic values for a single hue does exactly the same thing. Once we see the pattern of a single color changing value in regular perceptual steps (more on that in a moment), it sets up an expectation that this will continue. By mapping this expectation to a data series, it's easy to understand how that might highlight and enhance one's understanding of how the data is changing, too.
I have always hated the traditional color cubes for exactly the same reason as the professor: the units of control in the interface are wholly out of step with the units of perception. Move a little, and it is supposed to change just a little, but that is not what happens in a color cube!
Music has a similar problem, in that sound is not equally perceived across the range of possible combinations of vibrations. Early musicians invented "scales" of sounds, which are really just a sequence of sweet spots in these combinations that align with our own, internal "data" series -- the series of emotions and thought. When we hear a "sad" song, it literally makes us feel sad, the sound of sadness coincides so closely with the feeling. All of the sadness-inducing notes are collected together into a single, named collection called a "minor" key, something like a library from a programming perspective.
However, even with all the libraries of sound available, it was recognized very early on that the ranges are not mathematically perfect. Sound is composed by the summation of multiple vibrations, some of which cancel each other other, and others that emphasize each other. You have to "temper" the scales, that is, slightly tune them away from mathematical perfection, as you go up or down in pitch, in order for them to be equally perceived.
Re: (Score:1)
You don't 'temper' the scales to let them be more equally perceived. You temper the tuning to make it possible to play in different tonalities. As long as you only want to play in (for example) G major, and untemepered G major scale is going to sound better than a tempered one. But as soon as you want to play in a different tonality on an instrument perfetcly tuned to G it's going to be horribly out of tune.
Tempering evenly distributes the error over all tonalities, so it is least noticable. However it's st
I had this problem with heat maps (Score:2)
I programmed some heat maps once.
Grey-scale is no problem: you map the data range into 0-255 and you get nice grey scales.
Color was tricky.
If you just ramp the RGB values from red to green to blue, you don't get smooth color gradients.
Instead, you get all red up to the 1/3 point, and then all green from 1/3 to 2/3, and then all blue.
Right where the colors change over there are thin strips where one color shades into the next.
I had to twiddle the mapping function to stretch out those strips of shading in ord