25

I have just completed an article in which results are interpreted by means of 3D-plots; there are almost 15 plots in my article. Since plots could have a major impact on the reviewers’ decision, I am little worried about the colors used, which are actually very bright. Do the colors used in my plots really matter? If so, what type of colors should one should use in graphs?

  • 7
    Mandatory primer on choosing colormaps – user4512 Jun 26 '16 at 16:11
  • 10
    I wouldn't risk annoying a reviewer who'd discover that his gray-scale-printed copy is unreadable. – liori Jun 26 '16 at 17:23
  • 5
    Why? Sometimes distinct colors come out the same (or nearly the same) in gray scale. – GEdgar Jun 26 '16 at 18:17
  • 3
    @IgotiT: red and blue are notoriously hard to tell apart on a gray-scale print. – liori Jun 26 '16 at 18:33
  • 1
    Worth a read but by no means a dupe: academia.stackexchange.com/questions/44640/… – Chris H Jun 27 '16 at 8:08
50

Some people will tell you that colors shouldn't matter because only data should matter. Those people are wrong.

All human readers (including reviewers) find some graphical presentations easier to digest and comprehend than others, and the choices that matter include color. Of particular note, it is often important to consider color-blind readers, who are up to 10% of males, depending on population, (though many fewer females).

Color sets that work well for color-blind readers often work well for non-color-blind readers as well. You can find a number of good resources online to help you decide what colors to use, such as this "Color Universal Design" site, among many others.

|improve this answer|||||
  • well I agree that colors shouldn't matter. Unfortunately they do! :) – Ant Jun 26 '16 at 13:36
  • 4
    Another thing to keep in mind is that for heat maps and similar plots, gradients matter, and that the rainbow color map is notoriously bad for skewing how people perceive data. A good paper that touches on this is kennethmoreland.com/color-maps/ColorMapsExpanded.pdf – cartographer Jun 27 '16 at 3:56
  • 2
    Colours doesn't matter; data does. But when you present the data you want to make them easy to interpret - and colours, line styles, etc. are tools to enhance or bury the readability. So colours do matter. – Crowley Jun 27 '16 at 8:23
  • 8
    If someone tells you presentation doesn't matter, don't even give them a graph. Give them a huge table of numbers. – PyRulez Jun 27 '16 at 13:05
  • 1
    @PyRulez I've suffered through papers like that... – jakebeal Jun 27 '16 at 13:11
9

In very general terms bad design from trying too hard has more of a negative impact than lack of sophistication. So the best advice is to keep things as simple as possible.

Some basic guidelines are :

  • Keep to a consistent colour palette : it doesn't matter too much if this is bright primaries or more muted tones but be consistent. Primary RGB or CYM are both fine as are teal, beige and maroon but don't just mix and match and keep to one palette across a single document.
  • Think about what information you are trying to convey. If you need to use different colours to make superimposed plots distinct then use as many contrasting colours as necessary. If you are just making a simple side-by-side comparison of quantities then stick to one or two complimentary colours.
  • Don't use graphic effects (such as drop shadows and colour gradients) which are not strictly part of the information you are trying to convey.

The basic thing to take away is only use graphic elements and colours when you have a clear idea of how they help to make your information clearer. If your graphics work as monochrome flat images then stop there, only add additional complexity to graphics where it has a clear and definite purpose.

|improve this answer|||||
8

Yes, the choice of colours matters and http://colorbrewer2.org/ is a good place to find colour combinations that work well.

Also think about whether 3D plots are really the clearest way to present your data. Though occasionally unavoidable, there are usually better options. 3D plots work best in an online, interactive form, so if people are likely to read your paper as a PDF or in printed form, avoid them if you can. Can you show the same relationships in 2D form? Would a contour plot or heat map be sufficient? If so, one of those is probably a better option.

|improve this answer|||||
4

Recently, the default colormaps of several programs were changed:

  • Mathematica has a blue-orange-yellow colormap (before: white-blue)
  • Matlab (and this) has "Parula", which is a purple-blue-green-orange-yellow colormap (before: jet)
  • Matplotlib has "Viridis", which is purple-blue-green-yellow colormap (before: jet)

Unfortunately, I couldn't find specific information about Mathematica's changes (i.e., why did the choose this specific colormap). However, the linked posts for Matlab and Matplotlib provide some insight, which mainly agree to the answers provided here.

I highly recommend watching the video from the author of "Viridis": He explains, how he created this colormap, how you could create your own one and why "Parula" is a bad colormap.

Additionally, I'd like to point out that "Parula" is non-free according to a post of Steve Eddins from MathWorks:

The colormap is, however, MathWorks intellectual property, and it would not be appropriate or acceptable to copy or re-use it in non-MathWorks plotting tools

YMMV, but I recommend the "Viridis" colormap, because it enables gray-scale printing and is "compatible" to many kinds of colorblindness. You can find the RGB-colordata on github.

|improve this answer|||||
  • 1
    R also uses viridis – JenB Jun 28 '16 at 0:04
3

Of course, because a picture is worth a thousand words. A well-chosen bar-chart, a clear graph, add a lot to an already good paper.

Yet, since reviewers sometimes review "black-and-white" printed versions, it can be useful to combine colors with different line styles (solid, dashed) and markers (circles, squares), especially when curves cross.

Remember however that some journals require additional charges for color figures. If this is the case, and you cannot afford the cost, it would be a pity if you convince reviewers, but fail to convey your nice results to readers, because of poor graphics.

|improve this answer|||||

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.