When shown a bar graph that depicts average values, 1 in 5 people think that the bar's tip is the outer limit of the data, rather than the balanced center point (the average), as if every individual value was below-average. We call this error the "Bar-Tip Limit Error."
How could decades of graph interpretation research have missed such a severe error? Our experience in psychological measurement led us to ask a more concrete, direct, detailed question than prior work.
We developed a powerful new drawing-based measure of graph interpretation called Draw Datapoints on Graphs (DDoG). DDoG reveals correct and incorrect graph interpretation with unprecedented ease, building upon a rich tradition of drawing-based measurement in psychology.
As we write in the press release: "A major lesson from this work is that simplification in graph design can yield more confusion than clarification."
Broad take-home from our and their work: plots of summary statistics alone may yield frequent confusion and misinterpretation; the alternative of rich graphs that include individual-level data can be easy to make, easy to understand, engaging, and beautiful. #ShowTheData
An increasing number of great tools exist to easily create graphs that #ShowTheData. For example, our lab provides free, copy-paste-from-spreadsheet web apps at ShowMyData.org (built with #RStats and #RShiny).