Tableau Coxcomb Chart Template

 

Other than helping people with the calcs on the forums, I’ve never created a coxcomb chart in Tableau, but I recently had a use case for one and decided to build it from scratch. As a result, I’ve created a versatile plug-and-play template which I’ll be sharing here.

 

Note: I’m going to be going into some detail about coxcomb charts and how they work as well as my build process. If you’re comfortable with this, feel free to jump to the Template section to see how to use it. Or, if you feel like you can figure it out without reading that section, you can find the Excel and Tableau template files here: Coxcomb Template Files

 

What is a Coxcomb?

Coxcomb charts, also known as rose diagrams or polar area diagrams, were made famous by Florence Nightingale who used them to show the number of British soldiers who died each month during the Crimean War.

 

Image Source: Wikimedia Commons

 

The chart successfully showed that there were significantly more deaths due to disease than from wounds (right side of chart). Armed with this insight, she fought for better living conditions for troops, which helped to reduce the number of deaths from disease (left side).

 

Interestingly, she seems to have never called the chart a “coxcomb”. Rather, she used word “coxcomb” was to describe the books which contained the charts and that name was later given to the chart itself.

 

Coxcomb charts are a sort of modified pie chart where each slice/section represents some category or interval (in Nightingale’s chart, these represent a specific month). In a pie chart, measures are encoded using angles—larger values have larger angles. However, in coxcomb charts, all sections have equal angles and the measure is encoded using the area of the slice. Thus, unlike pie charts, the slices extend to different radiuses. A common misconception with coxcombs is that they are scaled by radius (linear scaling), but this is incorrect. The scaling by area leads to the different radiuses.

 

Each section of a coxcomb contains sub-sections encoded with different colors, each with their own areas. For example, Nightingale’s chart included sub-sections for deaths from wounds sustained in battle (red), disease (blue), and all other causes (black). Importantly, these sub-sections are not stacked as with a stacked bar chart. Rather, they overlap each other, all starting from a shared center point. Because of the overlapping nature of each sub-section, smaller sub-sections appear on top of other sub-sections so that they are always visible.

 

For a more detailed explanation of the chart and its history, I highly recommend the video by Robert Kosara, Florence Nightingale's Famous Rose Chart (aka "Coxcomb").

 

Building a Coxcomb

When I started building my first coxcomb, I made a huge mistake. Instead of taking the time to understand how they work, I just started building it. I assumed that a coxcomb was just a type of stacked radial bar chart, which led me to making two mistakes. First, I scaled each section linearly, rather than by area, which as detailed in the previous section, is inaccurate. Second, I stacked the sub-sections on top of each (like a stacked bar chart) rather starting each from a shared center point. These issues, of course, led to an inaccurate chart, but the second mistake, in particular, also made the build much more difficult than it needed to be. So, take this bit of advice—before building any non-standard chart, always be sure you have a clear understanding of how it works. 😉

 

Side note: Are you wondering exactly why coxcombs are encoded using area instead of radius? If so, I’d recommend another Robert Kosara video, Linear vs. Quadratic Scaling: How it Distorts Infographic Bar Charts, And How to Get It Right, which does a phenomenal job of explaining why we use area over radius.

 

Having realized my mistake, I returned to the chart and was able to modify it to work correctly.

 

 

I was very pleased with this and ready to share it with the world when I started thinking about whether or not there might be value to radial scaling and/or stacked sub-sections. I had recently created a data viz showing C-SPAN's Presidential Historians Survey and had used something similar to a coxcomb.

 

 

The chart shows the ten leadership qualities on which the historians scored presidents, each of which are measured on a scale of 0 to 100. In this case, it felt more natural for me to scale the sections by radius instead of area. This is particularly the case due to the fact that I added a grid to the background of each chart, showing intervals of 20.

 

 

With radial scaling, you can easily align each section to one of those grid lines. While that would have been possible scaling by radius, the linear distance between the gridlines would have decreased as the numbers grew, which I felt would be more confusing to viewers. Did I make the right choice here? I cannot be sure without a bit more research, but it felt like right approach in this case.

 

After doing a bit more digging for similar types of charts, I found a number of use cases (I’ll give some examples in a moment) where linear scaling and/or stacked sub-sections might make sense, so I decided to modify the template to provide you with a variety of options:

 

1) Scaling – Area or Radius

2) Stacking Option – Standard (no stacking) or Stacked

3) Whitespace – Specify a percentage of each slice which will be consumed by whitespace. This allows you to create some spacing between sections.

 

With these options, you can create four different charts:

 

1) Standard, Scaled by Area – A standard coxcomb chart

2) Standard, Scaled by Radius

3) Stacked, Scaled by Area

4) Stacked, Scaled by Radius – A radial stacked histogram bar chart

 

The Template

Like our other templates, this one includes two components—an Excel spreadsheet and a Tableau workbook. You can find them here: Coxcomb Template Files

 

The Excel spreadsheet has two sheets, Data and Model. Model is used to handle the data densification needed to draw the different sections of the chart. You don’t need to modify this sheet at all—just make sure it’s in your spreadsheet. The Data sheet is used to populate your data. It contains three columns. You can add more if needed, but these are the three that are required by the Tableau template. The columns are as follows:

 

Section – Dimension defining the sections (in Nightingale’s rose diagram, this would be month/year).

 

Sub-Section – Dimension defining the sub-sections (colors) within each section (in Nightingale’s rose diagram, this would be cause of death).

 

Measure – Measure to be plotted by the chart (in Nightingale’s rose diagram, this would be number of deaths).

 

 

Note: The template needs to include all combinations that will be shown on the chart. If there is a sub-section with no data, then make sure there is a row showing zero as the measure.

 

Once you have populated the Data sheet, then you need to connect it to Tableau. Using the Tableau template, edit the data source and connect it to your Excel file. The workbook should update automatically to reflect your data. Next, be sure to change the options to match your requirements. There are three parameters—Option, Scaling, and Whitespace. Once you’ve chosen the proper options, you can do whatever you like with the chart—change the colors, add filters, update tooltips, etc. just as you normally would.

 

Note: One common problem with any type of stacked chart is the inability to properly compare sections due to the lack of a common baseline. This template has the same problem when the stacked option is in use. To address this, I’ve used a sorting technique previously described by Steve Wexler and Ryan Sleeper. You can click on a sub-section to sort that sub-section to the middle. Additionally, in order to avoid that sub-section staying “selected”, I’ve leveraged the “fake highlight” trick described by Brian Moore.

 

 

Use Cases

In this section, I want to give a few example use cases outside of the standard coxcomb.

 

Pizza Charts

In my research, I came across “pizza charts”, which have been used quite effectively with football/soccer data, as illustrated by FiveThirtyEight, and the Football Slices twitter account. These charts generally only make sense when using a single sub-section. This is essentially the same chart as I used in my presidential ratings viz. We can use the template to create this by choosing to scale the chart by radius. As there is only one sub-section, the stacking option won’t matter.

 

 

Radial Stacked Histogram Bar Chart

In, 2018 I used Tableau to create a Wind Rose.

 

 

These charts are used by meteorologists to analyze wind patterns and show three key measures: direction, frequency, and speed. Direction is visualized by the angle of each bar, where North is 0° (the top), East is 90° (right), South is 180° (bottom), and West is 270° (left). Speed is visualized by the color of each section of the stacked bars. And Frequency is visualized by the length of each bar.

 

So, essentially, a wind rose is a radial stacked histogram bar chart (that’s a mouthful!), where the measure is encoded by radius and the sections are stacked on top of each other, not overlapping. We can create this using the template by choosing the Stacked option and scaling by Radius.

 

 

One thing I did with the original wind rose was add whitespace to create a bit of separation between each section, so if we change the whitespace option to 5%, we can do exactly that.

 

 

Or, if you’d like very thin sections, you could make it 50%

 

 

If you’re feeling really crazy, you can even set the whitespace to a negative percentage, which will create overlap on the sections. Now, I don’t know why anyone would ever want to do that, but it’s possible!

 

I don’t have specific use cases for every possible option, but if you happen to run into such a use case, the template should handle it without any problems.

 

Final Thoughts

Thanks so much for reading! As always, I want to make sure to stress that you should be very thoughtful about your chart choices. Coxcombs (and the related charts we’ve discussed here) have niche use cases, so be sure that you properly consider your data, your audience, and the questions you’re trying to answer before using this in your own work. And, once again, please be sure to view the videos from Robert Kosara because they do a fantastic job of describing this chart, its use cases, and the pitfalls you should try to avoid.

 

Ken Flerlage, January 18, 2022

Hire Us! | Twitter | LinkedIn | GitHub | Tableau Public