Reaction: Choosing Colors for Data Visualization

The blog gives nice examples as it goes along with the basic principles of color design, the problem of legibility, and some guidelines based on the principles. I agree that contrast and analogy are the basic principles where contrast draws attention, and analogy draws groups. The blog summarizes to assign color according to function, to use contrast to highlight, analogy to group, and control value(luminance) contrast for legibility.

It also gives a definition of hue, value, and chroma with sample images which makes it easier to understand. The terminology I use are hue, brightness, and saturation which confused me at first glance but was able to reorganize and match them.

The part unfamiliar was about "difference in value, formally specified as luminance contrast". If I had understood it right, it seems that value=brightness=luminance contrast. The gray text on a background varied from black to white gives a good example of why sufficient luminance contrast is necessary for better legibility. The layer effect caused by the variation in luminance shows another example of the benefit of luminance contrast. It would be a big advantage in visualization if we use luminance contrast in an appropriate way.

Regarding to Cynthia Brewer's "ColorBrewer" (http://colorbrewer2.org/), there is a limit to the number of color classes for the data set to 9. That is, ColorBrewer contains restriction if one wants to display the data with more than 10 colors. Even with 9 levels of color, it is still hard to distinguish some of the levels.