Source: http://www.physics.csbsju.edu/stats/display.html
Okay so you may have guessed, I’m not going to share secrets to using infographics to lie with your statistics. Instead I came across this cartoon on the web and it reminded me of the power information designers have over what story is being told of the data. Although there may be multiple stories within an individual dataset, it is always important to remember the way the data is presented to an audience can either help or hinder an accurate interpretation of it (even if you are not intentionally trying to lie with statistics).
Here are a couple things to check to make sure your data is being accurately represented in your infographic:
- Check proportion and scale. If you are using images or icons in place of bars, lines or pies, it may be easy to unintentionally skew proportion or scale of your data points.
- Be careful with 3-dimensional graphs. The angles or volumes used in these graphs can sometimes make it hard to accurately compare data.
- Make sure the appropriate information is included. Information overload can sometimes make it difficult to understand a graphic, but you also want to make sure your graphic is not too over simplified. Double check that critical information for understanding is included for your audience.
What are your biggest concerns or challenges when translating complex information to a graphic?
Thanks for the additional tips and comments. I am attending one of Edward Tufte’s course in a couple weeks. I look forward to sharing any additional insights that I learn at the training.
“Lies, damned lies, and statistics” is still one of my favorite phrases. Thanks for the post! Also, if you haven’t read any of Edward Tufte’s work, you’d really enjoy it. -Peter
You’ve hit my big three right on the head. (As a former data analyst, I’m well familiar with being pressured to lie through statistics.) There are two more that I have a beef with (both of which you actually allude to):
* Make sure your axes are labeled for bar graphs. The categories compared should be named, and the tick marks should be clearly labeled. This makes it harder to mislead (accidentally or intentionally) with scale. Consider the set 30.2, 30.5, 29.7; the graph’s going to look much different depending on whether the scale is 0-35 by 1 or 29.5-31 by 0.5.
For readers, this can be boiled down to: If the scale doesn’t start at 0, make sure they’re not trying to exaggerate differences between measurements. If the scale DOES start at 0, make sure they’re not trying to downplay differences.
* Don’t use the same graph to display different results. This is a personal peeve, I admit, but it really frustrates me when researchers try to save space by cramming two sets of unrelated results into the same graph. (Typically, these will have vertical axes on the left and right with different labels and tick marks, and sometimes horizontal axes on the top and bottom as well.) Like you said, it’s information overload, and it’s misleading at best to a layperson.
Good list!