Original

The most important aspect of a data visualization is honesty. The visual metaphor should be completely transparent. I just came across the following graphic of mortality statistics in England and Wales in 2010.

It does look pretty. And in terms of simple "show-the-data" honesty, it wins. If you calculate the deaths vs. area ratio for each circle is comes out consistent across data points. However, the basic conceit of a network is really not appropriate. Total deaths does not link or reference circulatory disease deaths, digestive disease deaths, etc..., but instead contains each of the more specific causes. This is a hierarchical data structure. We can represent this hierarchical structure naturally using a tree map or mosaic graph.

Click through for full-sized version with labels and counts.

Image © 2011 Guardian News and Media Limited.

Redesign

Here's my re-implementation using the exact same data, but displayed using d3.js's Treemap function. The area of a box represents number of deaths in 2010. Boxes are grouped into bigger boxes according to cause of death.

Click through for a Javascript version that resizes to the window. In the Guardian's version it was hard to tell the relative contributions of different causes. Here, comparisons are easy. For example, one can immediately see that cancer and heart disease are the predominant killers. Additionally, this version has the advantage of a higher data-to-ink ratio; there isn't so much wasted white space. I could imagine improving this further by shading light vs. dark based on change from previous year.

Data is for all deaths in England and Wales in 2010, originally compiled by the Office for National Statistics.