News and Notes from PolicyViz - Issue #18

Jun 21, 2022

Summer is in full swing and I'm just returning from a week away! I hope you will get some time to rest and relax this summer as well!

Some colleagues recently asked me to help create a visualization of the same set of questions asked of two groups. The data for each question was a percentage between 0% and 100% and they wanted to show both the raw value and the rank of each. They originally created two bar charts, which accomplishes the first goal, but makes the second goal much more difficult.

I played around with a number of things, like a dot plot and a slope chart. Because I can't show you the actual data, I'm just using "Option X" for each label, but in the original, each was a long question from the actual survey.

Even though I colored the lines differently depending on whether the rank increased (blue), decreased (green), or stayed the same (yellow), this was just a little too hard to read, mainly because--I think--the lines are showing changes in value, not rank.

Anyways, I then thought about making two bar charts and just linking the bars with some lines. This was my first attempt:

Bar charts with connecting lines of different lengths

I liked this as a starting point, but someone in my Twitter feed noted that the slope of the line is affected by the length of the bars. That is, Option #16 and Option #18 both move down one place in the relative rankings, but the slopes of those two blue lines are different because of the length of the blue bars.

Then, someone else reminded me of this piece from Nathan Yau over at Flowing Data:

Simran Parwani @simranparwani

@jschwabish this makes me think of a recent project from @flowingdata : https://t.co/q6SHSrnCH9

In that version, Nathan showed rates of social media usage by age group. Instead of having the bars stop at their end point, he created a little outline around each so that the length of each is the same--a 'unfilled tank'-look, if you will.

I tried this with my data:

This, I think, is better. Maybe I don't need the bars to go all the way to 100%, but I like the general approach. I was able to do this in Excel by combining bar charts and scatterplots. (I may pull together a video tutorial at some point this summer.)

In the end, my colleagues didn't use any of these versions! Instead, we iterated on one where there is one set of Options listed from greatest to smallest along the first survey question. The bars next to them are listed in the same way with a different color used to highlight the series that changed by at least 10 percentage points.

It's a somewhat simpler approach but perhaps gets to their point a little bit faster and easier. It's also a little easier to create using Excel's conditional formatting menu.

As always, what is most important is your audience and the best way to get your audience to understand the story you want to tell.

Thanks,

Jon