Happy Holidays from PolicyViz - Issue #28

Dec 20, 2022

Hi all,

It's the last newsletter of 2022! Phew, we made it! It's been quite a year and I'm fortunate to be taking the next couple of weeks off, so don't expect too much from me on social media or elsewhere.

In case you missed it, another one of the Do No Harm Guide reports was released last week, this one on accessibility in the data and data visualization fields. Accessibility took a front seat in the dataviz field this year, I am so proud of this report, which I hope can be a cornerstone document for everyone seeking to make their work better and useful to more and more people.

The report consists of nine essays (plus an introduction) from folks all around the accessibility world and covers a range of issues from alt text to screen readers to organizational structures. I hope you'll take a read because this Do No Harm Guide: Centering Accessibility in Data Visualization is an absolute fabulous resource.

One last thing. In case you didn't see, Twitter is closing the Revue newsletter service come mid-January. I'm pretty bummed because this service is free and pretty easy to use. I'm exploring other options, most likely Substack. I'll move your subscription over to whatever comes next, but I appreciate you taking the time to read this newsletter and am grateful for your support.

I hope you have a wonderful holiday season! I hope you are able to spend some time away with family and friends, staying happy and healthy. Thank you again for all of your support at PolicyViz, either by reading this newsletter, visiting the website, listening to the podcast, or chatting with me on Twitter and elsewhere.

Until next year,

Jon

DRAFT POST: Avoiding the Dual Axis Chart, Part II

A few months ago, I wrote a post about why and how to avoid dual axis charts. In that post, I focused exclusively on dual axis charts in which both series are lines, that is, showing changes over time. But there are other kinds of dual axis charts—for example, charts that show two categories on two separate axes, and that is the kind of dual axis chart I want to focus on in this post, and again provide some alternative approaches.

Let’s start with a simple example, something that you may have come across at various times. The graph below shows two metrics: life expectancy (measured in years) and health expenditures (measured as a percentage of GDP). The life expectancy metric is encoded as vertical bars associated with the left vertical axis. Health expenditures are shown as dots and associated with the right vertical axis.

As I argued in the previous post, these kind of dual axis charts can create several perceptual challenges:

it’s not quite clear which series correspond to which axis
the axis labels and gridlines may not match up (notice the percentage labels on the right vertical axis don’t sit on the gridlines)
our eyes may move to intersections (or separations) in the two graph objects (i.e., bars and dots) that may not correspond to important differences.
I’ll add one more here—they are just darn confusing. Too many bars and dots and overlaps and whatnot.

Alternative Approaches

The previous approach suggested several alternatives to the dual axis line chart problem, including arranging separate graphs vertically or horizontally, or alternative graph types like the connected scatterplot. In addition to simply leaving the two series on two separate graphs, there are at least three different alternative approaches.

Scatterplot

If the point of this kind of dual axis chart is to look at the relationship between the two series, how about a scatterplot? In this case, I put health expenditures as a share of GDP on the horizontal axis and life expectancy on the vertical axis. I’m not sure if the scatterplot is more or less familiar than the dual axis chart, but I do find it a bit easier to read. It's also worth pointing out that Michael Friendly recently noted that Playfair's famous dual axis chart might have been the best chart at the time because the scatterplot had not yet been invented.

The other advantage is that the scatterplot can include additional data such as sizing the circles to a third variable or adding additional years to the series, such as a scatterplot with trails, which is really just a connected scatterplot. The graph below is from Our World in Data and has the same organization as the graph above but includes additional years of data.

Marimekko/Mosaic

Another alternative is to use a Marimekko chart, which scales the widths of bars in a vertical bar chart to a second variable. In this case, life expectancy is shown along the vertical axis and the horizontal axis shows health expenditures as a share of total health expenditures among these countries. (I learned how to create a Marimekko/Mosaic chart from this short video from The Information Lab.)

Parallel Coordinates Plot/Slope Chart

Yet one more alternative is to try a parallel coordinates plot. Now, personally, I’m not a huge fan of the standard parallel coordinates plots you might see in a Google search with tons of lines and multiple vertical axes. But with just two vertical axes, I think they can be okay; these data may not be the best example for this kind of chart, but hopefully you get the idea.

I added the words “Slope Chart” after the forward slash because many people would probably call this chart a slope chart. Personally, I reserve that term for a chart of the same format but when used to show change over time. The form is basically the same. (I learned how to create a Marimekko/Mosaic chart from this short video from The Information Lab.)

Conclusion

I’ll end this post the way I ended the first one: I hope I’ve demonstrated how dual axis charts can be troublesome. They can be confusing, difficult to read, and even harder to figure out what the point of the chart is supposed to be. Some of the alternatives here—not to mention the tried-and-true strategy of just using separate graphs—are hopefully strategies you can try in your own work.

Please let me know if you have thoughts on this blog post. I post drafts here and then put them on my site about a week later.