Radar charts are much maligned in the world of data vis. Every other explanation of their workings comes with a disclaimer explaining why “you probably want to use a bar chart”.
Their problems include:
- The shape itself doesn’t mean much: depending on which order you place the variables around the circle, the shape can change a lot, turning from a star to a blob and back again.
- The volume of the shape is not a good indicator of anything, since the volume increases quadratically with only linear increases in the variables, as you approach the outside of the circle.
- There’s no sensible way to add meaning to the order of variables – unlike a bar chart that can be ordered from highest to lowest to make understanding easy, a circle is non-hierarchical.
Overall, there’s a lot of ‘meaningless information’ in radar charts. The exact form of the shape, the volume, and the ordering of variables are all mostly noise.
But in the case of showing media trends and biases surrounding the Israel Folau case recently, we decided to use radar charts. We had some good reasons – but first a little background.
There are a lot of issues at play in the Israel Folau case, and different media outlets seem to focus on quite different aspects of the case. Homophobia looms large, but different media chose to focus on other facets of the case more than others – political correctness and religious freedom, for example. For an article published on Streem, we gave every issue its own radar chart, with a central ring representing the ‘average’ amount of coverage of a given issue. Outlets with above average coverage of an issue stick out.
Here’s some of the results, from searching 2772 articles for themes and keywords:
Some things are clear – The Australian has an outsized interest in political correctness, for example. Yahoo!, interestingly, provided the most discussion of homophobia, while the Australian Financial review was predictably focused on the business side of things – the implications for rugby sponsor Qantas. We’ll let you draw your own conclusions from the graphs themselves:
Here’s why we think radar charts are actually a pretty good idea for showing bias:
- In this case, it’s a good metaphor. If you stack all your cargo on one side of a boat, it will tip over. If you distribute it evenly, the boat will be balanced. As a physical metaphor, we instinctively understand that a centred, regular shape is ‘balanced’, and an off-centre shape is unbalanced or ‘biased’. So a radar chart inherently lends itself to the idea of measuring balance and bias (bias simply in the sense of focusing a lot on one issue). Similarly even if a star is fairly symmetrical, we understand that it’s not an ‘even’ shape. Well-rounded coverage of an issue is well… rounded. Metaphorically, it’s a nice round circle.
- Bar charts don’t imply ‘bias’. Another way to show the same data would have been using a bar chart. Draw a horizontal line at the average, and any media significantly above or below that would be exhibiting bias. But the metaphor is all wrong. In the context of height, we generally understand that ‘up is good, down is bad’ (or sometimes vice versa). The idea that bars should converge around a mean height is not easy to grasp.
So while it’s true that radar charts are hard to understand in a lot of cases, we think there’s a use case for every chart. Radar charts to show media focus just works. Even the circular form of the radar chart also implies focus – is the media coverage focused (converging on an average centre), or does the coverage range widely by outlet?
Do you agree? Disagree? Leave us a comment here or on Twitter.
Or for anything else, get in contact.
