Which Color for your Figure? 

 ever wondered about what would be the best color for your graphs? While common in the sciences, it may be fair to say that the use of color in graphs is still under-appreciated in many social science fields. Colors can be a every effective tool to  visualize data in many forms, because color is essentially a 3-d concept: 

 - hue (red, green, blue) 
- value/lightness: (light vs. dark) 
- saturation/chroma (dull vs. vivid) 

 From my limited understanding of this topic, not much scientific knowlegde exists about how color is best used. However, a few general principles have emerged from the literature. For example, sequential information (ordering) is often best indicated through distinction in lightness. The tricky part here is that indicating sequence with colors requires the viewer to remember the color ordering. A small number of colors should be used. One principle that is sometimes advocated is the use of a neutral color midpoint, that makes sense when there is a "natural" midpoint in the data. If so, you may want to distinguish above and below the midpoint, and use dark color1 -> light color1 -> white -> light color2 -> dark color2 (e.g., dark blue to dark red) . If no natural midpoint exists, one option is to use a single hue and just vary lightness (e.g., white/pink to dark red). Another idea is that categorical distinctions are best indicated through hue (e.g., red=higher than average, blue=lower than average). Read Edward Tufte and the cites therein for more ideas on the use of color. In addition, a nice online tool that helps you choose color in a principled way is  ColorBrewer , a website definitely worth a visit. Many of the color schemes advocated there are also available in R in the ColorBrewer {RColorBrewer} library. Good luck!