donut chart variant: howto

February 2, 2016 11:45 AM

I need to know the name of this doughnut chart variant and also a means to replicate.

My instinct is to call it a stacked doughnut chart: http://prntscr.com/9y73a5. We have a similar data set, with progressive groups increasing the area, but have not found an easy way to replicate it using excel, spss, or online tools. Thanks much.

posted by cgk to Computers & Internet (8 answers total)

Best answer: Looks to me like a bespoke chart created by a graphic designer. Seems unlikely that there's a tool to generate them, as it's the sort of chart that will only work for data sets with quite specific characteristics.

You can probably replicate it by exporting individual charts to PDF and then using Illustrator or Inkscape or similar to do the stacking part.
posted by pipeski at 11:54 AM on February 2, 2016

It would be easy to do in matplotlib, which is a graphing library for Python. Here's a gallery example:

http://matplotlib.org/examples/pie_and_polar_charts/polar_bar_demo.html

You'd set the bottom of each bar to something larger than 0.0, and put in location and width so they don't overlap, but the arithmetic for that isn't hard.

Macs come with Python, and matplotlib is free, but you'd have to install matplotlib.
posted by clew at 12:05 PM on February 2, 2016

To me, that looks like a regular old doughnut chart that's just been altered for decorative effect. Data vis designers generally recommend against this kind of thing, FWIW, because it can throw off the reader's perception of how big the segment really is. But if you want to do it, you can probably save an Excel graph as a PDF and noodle with it in a vector image editor (Illustrator / Inkscape).

A variant where the size of the portion actually does convey something about the data is called a coxcomb chart.
posted by fifthpocket at 12:07 PM on February 2, 2016 [2 favorites]

I'm pretty sure your example graph wasn't made with a tool, or at least not a very good one. The colored wedges of doughnut do not correspond exactly with their labels; the section for 9th graders subtends about 100°, which is closer to 28% of a circle. The section for 11th graders seems closer (125°).
posted by aubilenon at 12:16 PM on February 2, 2016

Apologies for not answering the question directly, but as a data guy, I second fifthpocket that this is not a good way to represent the data in an unbiased manner (this is the kind of nonsense that gets people railing against how stats are presented in the mass media). Your readers will intuitively gauge the relative sizes of each proportion by the area they cover rather than the angle that actually reflects the proportion, and this is the type of graph you'd use if you wanted to exaggerate the relative size of the 26% proportion.

Logistically you're also going to be in a pickle if you're presenting the results of multiple questions using the same format and breakdowns -- (using the posted example) what happens when in the question after this the 9th-grade proportion is bigger than the 11th grade proportion? You can't just reverse the order in which the choices are nested within that question's graph, since then it becomes even more misleading.
posted by Doofus Magoo at 1:16 PM on February 2, 2016

I think it's actually kind of awful presentation, WRT the 11th grader data. When I first looked I though they were about 10%, not 34%.
posted by achrise at 1:24 PM on February 2, 2016

Response by poster: Thanks everyone. Expected disclaimer: this is a request from much higher up and they like the graph format -- certainly something I will never use again.

I informed them that is was probably a custom graphic job, they decided to go with a concentric doughnut graph. I appreciate everyone taking the time to help me out.
posted by cgk at 4:20 PM on February 2, 2016 [1 favorite]

I’d just like to drop my opinion in that, while the chart is cute and artsy, it’s also bad and confusing.

The 34% of 11th graders is shown by a light blue wedge that only takes up maybe 8% of the donut.

Looking longer than one should have to, I guess they are somehow trying to lump the 11th-graders in with the dark blue 26% that indicates the 9th-graders...
like adding the 11th-graders to the 26% of 9th-graders to make 34%.

But that still doesn’t make sense; the dark blue wedge should be it’s own 34%, unless they are somehow trying to put all the high-schoolers into one wedge or something... it’s... it’s just a badly done chart, putting form (cuteness) above function (how percentages work).

A chart like this could be done well—just make a pie chart and then drop a white circle into the center to make it a donut—but this one is just bad.
posted by blueberry at 6:49 AM on February 3, 2016 [1 favorite]