At PresentationGuru on January 17, 2019 there was an article by Rosie Hoyland titled The best way to protect yourself from misleading graphs. It linked to a four-minute TED-Ed YouTube video by Lea Gaslowitz from July 6, 2017 titled How to spot a misleading graph, and also warned that:
“Even careless handling of the data can send a distorted message.”
I found a good example of careless handling in a long, otherwise decent article by Michael Smith at SlideHeroes titled CONQUER THE FEAR! 8 Steps for Controlling Public Speaking Anxiety, which contains the bar chart shown above that displays results from the 2014 Chapman Survey of American Fears. The top blue bar indicates that 29% of Americans fear public speaking.
Compare that with my bar chart showing 25.3% (for the same sum of Very Afraid and Afraid) from a blog post on October 29, 2014 titled Chapman Survey on American Fears includes both zombies and ghosts, and the Washington Post version from an October 30, 2014 article titled America’s top fears: Public speaking, heights, and bugs.
How did Mr. Smith wind up with an answer that was inflated 4% higher than mine and the Washington Post? Look at the rather confusing data table from the survey for this item. As shown above, the first row in the data table is for Refused – people who didn’t answer because they don’t know. But he read and used 29.3% - the Cumulative Percent (right column) from the third row, which wrongly includes that 4.0 percent from the Refused (or Don’t Know) category. Really 8.8% were Very Afraid, and 16.5% were Afraid, which adds up to 25.3%. Oops! Every item in Mr. Smith’s chart includes that same mistake.
A Dilbert cartoon on January 30, 2019 described another type of misleading graph:
Presenter: As you can see from this chart, our product has been rated number one for six years in a row.
Dilbert: Why does you chart stop four years ago?
Presenter: I’ll bet you don’t get invited to a lot of parties.
Dilbert: That’s just a lucky guess.
The cartoon of a man with a pump was adapted from one at Openclipart.