This video is part of the How to Avoid These Data Visualization Mistakes series, presented by Naomi B. Robbins, Data Visualization Expert at NBR.
Transcript:
The next few slides I'll be talking about color. There are three properties of color that define it: Hue, saturation and lightness. Hue is wha ...
This video is part of the How to Avoid These Data Visualization Mistakes series, presented by Naomi B. Robbins, Data Visualization Expert at NBR.
Transcript:
The next few slides I'll be talking about color. There are three properties of color that define it: Hue, saturation and lightness. Hue is what we mean when we ask a two-year-old what color is this? In the left of this chart, the hue was blue in the right, the hue is orange.
Saturation is sort of the purity of the color. As you can see, as the saturation increases, the orange is a purer orange. The blue is a purer blue. As the saturation decreases, the colors get grayer and lightness/darkness, they're whiter when they're lighter, and blacker when they're darker. I wish my only improvement this slide would be to say darkness/lightness so that it went in the same order as the figure goes, but that's what we mean by hue, saturation and lightness.
This was done by Cindy Brewer, and she discusses three types of schemes for choosing colors. In sequential scheme you have one hue with varying saturation and lightness, and that's very good for quantitative data. You go from lighter to darker. Qualitative schemes are different hues with the same saturation and lightness, which is very good for categorical data.
Things, if you have three countries, you don't want to make one more important than the other by having it more saturated or so that different hues at the same saturation is good for qualitative. And diverging, you have a neutral color in the middle going out in two directions. For example, with the Likert scale, you might have no opinion as neutral, and then agreement going out in one color, and disagreement going out in the other.
It's important to use the right color scheme. Here is a map. It comes from the Journal of American Water Resources Association, and you can see a big divide in the middle of the country where the yellow and green, or at least it looks like a big divide. Now all I'm going to do is blow up this scale on the bottom so you can see it more clearly, and you can see that the green is .7 to .79, the yellow is .8 to .8. They're not that different. It looks like there's this big difference, but they're adjacent values.
What I've done here is just do it in gray scale, so you can see .7, .8 are very different, but they're adjacent, and you can see that going from 0 to 1.29, we go all over the place in colors we get darker, lighter, darker, lighter so that the impression you got from the qualitative scheme is completely false.
And here it is in a sequential scheme, and you don't see that divide at all. The slight difference of color you see is quite a bit further west than the divide we saw before. And if you look there, you can see that the colors are sequential. We go from light to dark.
A major problem in charts done in almost any field, not just life science, is that they don't consider people with color vision deficiencies. Lots and lots of books I've read, and lots and lots of speakers I've heard have said, just avoid red and green. Just avoiding red and green does not solve the problem. Most people who have color problems can distinguish between a very dark red and a light green, or a very dark green and a light red. Lightness has a lot to do with it, and here I'm going to show you several figures that are orange and green.
Orange and green was the latest fashion in recent years. This is from the Census, and here I've done a simulation of how people with color problems, this one is protanope, which have people who have problems with their red cone, and you can see how they see it. This one is dueteranopes, which are a problem with their green cone, and again we see orange and green on the left. Someone with problems might see it as we see on the right.
When I draw a chart, I always run it through a color vision simulator. I used to use Vischeck. It is not available anymore. I now use Color Oracle, and I've discovered that Graph Pad people use Color Oracle. So I highly recommend running your figure through something like Color Oracle. That's good for, I believe both PC and Mac. There are a lot of them for a Mac. I happen to have a PC and don't know that many.
Using gradient backgrounds can be a problem. If you look at the left, I have a gray square. All the squares on the right are exactly that color, but they look darker on a light background. They look lighter on a dark background. So gradient backgrounds are a problem. This shows you why.