How to colour your heatmaps.
- Top-left is not Tufte-compliant because primary colours occupy too much of the space (it’s distracting).
- Top-left also uses colours that do not grade across our perceptual space. (Although hues do grade across wavelengths of light smoothly, we don’t perceive it that way.)
- Topographical maps that use green, brown, and yellow likewise do not grade across perceptive colour space appropriately.
- Top-right is fine but perhaps a little bland. A topographical map with a lot of hills and valleys would benefit more from this than one trying to show finer detail. (My intended application—overlaying two elliptical single-peaked utility functions—would have a hard time with such an approach.)
@hadley recommended this paper to me. I was asking how to select colours to represent level curves (isoclines / isoutility curves / etc) on a 2-D plot. (I.e., how to plot a ℝ²→ℝ¹ function effectively with colours other than greyscale+red.)
Big-money quote from Zeileis, Hornik, and Murrell:
It has been hypothesised that human vision evolved in three stages:
- perception of light/dark contrasts;
- yellow/blue contrasts (usually associated with our notion of warm/cold);
- green/red contrasts (helpful for assessing the ripeness of fruit).
….. The subjective experience of [colour, however,] is less well understood.
Wikipedia pages to read:
- RGB (old news)
- HSV (old news)
- HCL = Hue Chroma Lightness
Book to read: