[This is a guest post by Severino Ribecca*, as part of a series dedicated to each individual kind of chart that he has read into as part of his main research project.]
This assortment of multi-coloured blobs scattered across grid is known as a Bubble Chart. Ideal for multi-variable data, this type of chart resembles a combination of both a Scatterplot and a Proportional Area Chart. Bubble Charts are great for when you have three or more variables in your data.
Much like the Scatterplot, Bubble Charts use a Cartesian coordinate system to plot dots onto a grid where the X & Y axis are both separate variables. However, Bubble Charts takes it a step further and assigns another variable the graph by varying the size of each dot. Each of these dots are also assigned with a label or category. Colour and shading can also be used to distinguish between categories or be used to represent yet another variable in the data, through different shades.
Typically, Bubble Charts are used to compare and show the relationships between the bubbles by the use of positioning and proportions. The overall pattern created by a Bubble Chart can also be used to find correlations, make in the same way a Scatterplot would.
Above is a colourful example of an interactive Bubble Chart, which displays the scientific evidence on popular health supplements. The higher up on the graph a bubble is, the stronger evidence there is to support it. The x-axis is used to arrange the bubbles either alphabetically, by popular interest or by scientific interest. Bubble size and colour can be changed as well via the menu options.
Another example from Gapminder is a Bubble Chart that plots each country in the World in terms of it’s life expectancy and GDP per capita in 2010. You can see a fairly strong, positive correlation between the two variables, where countries that have a higher GDP per capita generally have a longer life expectancy. The area of each circle is also used to show each country’s population and colour has been used to divide bubbles up based on geographical regions.
While I was unable to find the origin of Bubble Charts, the earliest resemblance of one I could find is the Hertzsprung-Russel Diagram (H-R diagram) invented in around 1911 – 1913 by Ejnar Hertzsprung and Henry Norris Russell. I first came across the H-R diagram back in school when I was studying cosmology as a sub-subject in physics.
The H-R diagram has a lot of variations, but typically stars are plotted against star surface temperature on one axis and absolute magnitude on the other. The size (not mass) of each star can also be displayed by the star’s area on the graph. The development of the H-R diagram helped provide an entirely new way of looking the at star, which laid the groundwork for modern stellar physics. From estimating the masses of stars, astronomers discovered a relationship between the mass of “main sequence” stars from looking on its position of a H-R diagram.
Animated Bubble Charts
Animation can be used in Bubble Charts to show the data changing over time. Below are a couple of videos presented by Hans Rosling that use time and animation in Bubble Chart to tell a story:
200 Countries, 200 Years, 4 Minutes – The Joy of Stats
Showing the increase in life expectancy and wealth across the countries of the world over the last 200 years.
Religions and babies
Here, Rosling shows that contrary to popular belief, religions are not the factor that influences the birth-rate in countries around the world.
Problems with Bubble Charts
Just as you do with all charts, Bubble Charts have their own flaws and limitations. Having too many bubbles can make the chart hard to read, so the data size capacity is limited. This problem can be somewhat remedied by interactivity, by having an option to reorganise the bubbles or to filter out grouped categories. Also the use of transparent bubbles can prevent bubbles from being hidden and markers can be displayed in centre to enable the reader to more accurately determine the exact location of the data point:
Like with Proportional Area Charts, the sizes of the circles need to be drawn based off the circle’s area, not its radius or diameter. Not only will the size of the circles change exponentially, but this will lead to misinterpretations by the human visual system.
The next post will be looking into Choropleth maps.
Research & Further Reading:
• Bubble Chart Reference Page – The Data Visualisation Catalogue
• Wikipedia entry on Bubble Charts
• Gapminder World – Interactive Bubble Chart
• Bubble Charts – fusioncharts.com
• The Power of Bubble Charts – The Plotly Blog
*Severino Ribecca is a British graphic and information designer interested in data visualization. Currently he’s building an online library of different information visualization methods called The Data Visualisation Catalogue. You can follow the project’s updates on Twitter (@dataviz_catalog) and support further developments on the Patreon Page.