Don’t get caught lying with data – check out Tufte’s Lie Factor

Greg Nelson Analytics training

The whole point of depicting data and information in a graph is that someone can visually see what’s going on, what the relationship is between variables. However, it is easy to take for granted not only how that information is displayed, but how it is scaled. Obviously, when you look at a graph you have a first impression of the story the data is telling. For example, check out this graph for the 2008 election. Clinton is blowing Obama out of the water, right?

Well not exactly, a different graph showed a much closer race between the presidential candidates.

Simply put, Tufte’s Lie Factor is a formula that helps one determine how visually accurate the data being presented are. The image should be proportional to the information. The formula is (the size of effect shown in the graph)/(size of the effect in data). Ideally you want this ratio to be in between 0.95 and 1.05.

Looking back at our example graphs, Clinton has about 11,750 votes and Obama has about 10,800. The size of the effect in this data is 0.08. This is calculated by taking the first value, of 11750 and subtracting the second value of 10800 and divide that number by the first value. However the size of the effect in the first graph is 1.5 which gives us a lie factor of 18.75. WHOA!!! Way off. 

We love looking through graphs to see what sort of mistakes, intentional or unintentional, were made. Let us know if you spot any and send them our way. And if you are looking to learn more about data visualization, head on over to our academy and check out our Data Viz course.

Read the source article at DataVis.ca