This video is part of the How to Avoid Common Data Visualization Mistakes series, presented by Naomi B. Robbins, Data Visualization Expert at NBR.
Transcript:
Thank you. I'm here this afternoon to discuss how to avoid some common graphical mistakes. Here, we have the one, two, three, four. We have t ...
This video is part of the How to Avoid Common Data Visualization Mistakes series, presented by Naomi B. Robbins, Data Visualization Expert at NBR.
Transcript:
Thank you. I'm here this afternoon to discuss how to avoid some common graphical mistakes. Here, we have the one, two, three, four. We have the row, the column, and these three things that I've called S1, S2, S3, S4. Every one of those rows and columns is a permutation of two, four, six, eight. If you look at the four bars with the red arrow going through them, they are all six.
They don't look the same because of the way they're laid out, and if we look at three, you don't see the two because it's behind the tall ones, so even if you really have three dimensions, these 3D bar charts are confusing. What would I do instead of this? I would do four separate charts. I would fix one of the variables. In this case, I fixed the S's and then I'd show the other two variables for each value of the S's.
Here's another case where we have unnecessary dimensions. We know one thing about the data, its value. You don't need a two dimensional figure, like a circle, to show one dimension, and what's more, these are very confusing, because you don't know what is encoding the value. Is it the diameter or is it the area?
Here are two figures. They look very similar. We see a bunch of circles in size order in both of them, but in the top one, the candy, the data is encoded in the diameter. In the bottom one, it's encoded in the area. Which do you think makes more sense? We see the area, so that the bottom one is a much better figure than the top, but because people don't know which is which, it doesn't help.
This graph is blatantly deceptive. I found it in a corporate annual report. Let's look at the dividends per share, the bottom one, we go from 1996 to 2000 and we go from 16 pence to 21 pence. 21 looks many, many times 16; it's just a blatant lie. If you drew it to scale, it's increasing, but, not nearly as fast to see other one looked. So, we've seen lots of examples of using unnecessary dimensions and they all made the charts mislead.
If you're doing a bar graph, do a simple 2D bar graph. Make it beautiful by your choice of the width of the bar, the spacing between bars, pick a pretty color, but don't try to wow your audience with an unnecessary dimension.