Common Data Visualization Mistakes Part 3: Comparisons with Different Scales

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:

Comparison with different scales. Let's go back to our New England Journal of Medicine. I showed you just one of these before, but the way it was done, we have men and women, one on top of the other, which invites comparison of men and women, but look carefully. The women go from 0.6 to 2.2. The men go from 0.6 to 2.8. That's different.

Now I want you to look at the distance between 0.6 and 0.8 on the bottom and the top. It's much bigger on the bottom than it is on the top. You can't compare those two when their scales are different, you get a distorted comparison. Well there it was the scales that were inconsistent.

Here it's the color. On the left, 2010 is green. The other years are sort of cyan. If we go to the right, it's 2012 that's green. 2010 is the Cyan. We've changed what the colors represent from the left to the right.

Another one. Here are two pie charts. If we look at online on the right, it's green. If we go to green on the left, it's something different and online is gray. If you look at one of the labels you say, "Oh, green means..." It doesn't mean that on the other.

Here's a different sociology textbook. We see that each bar is a different color. There's absolutely no need to make each bar a different color, unless for example, you're interested in a particular country. In this book, lots of the examples have countries. For example, suppose I'm interested in Sweden. I look on the left and its gold. I come to the right, I don't see gold, but wait, here's Sweden, it's purple here. If you're going to use colors, have a meaning for them. Would you think of writing sentences with each word a different color? Then why make each number a different color? Why not show the same respect for numbers that we show for words?