I just saw this interesting series of charts over on Business Week (click image to view full-size). It just goes to show that you really can force data to give you the answers you want. All you need is one leading question, and 2 simple charts, and voila you can pretty much prove anything. 

Correlation may not imply causation, but it sure can help us insinuate it.

It’s pretty scary really… this simple exercise shows humorous examples of how this technique can be used to manipulate data, but how often is it used in the media or political campaigns to confuse or brainwash? A truly worrying example of how data can be used for evil. 

View full-size graphs and article on Business Week

I saw an astonishing sight a minute ago - this graphic (click to see full size and origin):

At first glance this appears to be your standard polar area diagram, or more commonly referred to, radial bar chart… however it has taken a useful visualisation method and used it in a way it should not be used. I see what they’re trying to do, but the method they have used to convey this information is completely inaccurate and misleading. Radial bar charts should never be used to show different statistics from different data sets. 

For example:

88% of people are aware of Facebook

Does not directly correspond with:

30% log in with a mobile device

These facts therefore, cannot be displayed on the same graph / chart. It doesn’t make sense. You wouldn’t display these individual statistics on the same standard bar chart, so there is no reason why you would display them on a radial bar chart.

Polar area diagrams, correctly used, primarily display changes over time, as shown in this classic by none other than Florence Nightingale (click to view full size):

Here Nightingale illustrates the causes of mortality over time, with each colour representing a different cause, and each section of the ‘pie’ (if you will) representing a different month. Data from one set only is used to show the difference / change over a period of time. This logically makes sense, and isn’t misleading. The viewer can instantly see that in certain months more died than others, and what the primary cause was.

There are some drawbacks from using this method as some would argue it would just be easier to display the information in a standard bar chart, and they likely are right as data can be more easily extracted from information displayed in a linear, as opposed to radial, fashion. 

But primarily from a visual standpoint, the radial polar chart is much more striking. And if you have a large change over a small amount of time, the effect can be quite dramatic if presented using this method. 

There are many for and against such methods, and my aim wasn’t to discuss this here, but merely to explain how such a method should and should not be used. Unfortunately the top graphic used the method ineffectively rendering the chart useless. The facts here, would have almost been more beneficial to me and easier to read had they merely been presented in a list or a series of smaller separate charts.