Anchors and ratios to quantify and explain y-axis distortion effects in graphs

Area of Science:

• Data visualization
• Cognitive psychology
• Information design

Background:

• Data visualizations are prevalent in scientific and public communications.
• Y-axis range manipulation is a common method for distorting graph perception.
• Truncating the y-axis (starting above zero) can inflate perceived differences in bar graphs.

Purpose of the Study:

• To define 'anchors' as perceptual information for explaining bar graph distortions.
• To investigate the existence and impact of upper y-axis truncation.
• To examine the effects of y-axis truncation and expansion on perceived differences across various graph types.

Main Methods:

• Four studies were conducted to test y-axis distortions on different graph types.
• Experiments involved manipulating y-axis ranges (truncation and expansion) and graph formats (bar, dot, line).
• Observer judgments of value differences were recorded and analyzed in relation to graph distortions.

Main Results:

• Both lower and upper y-axis truncation increase perceived value differences; truncation has a larger effect.
• Y-axis expansion decreases perceived value differences compared to standard scales.
• Bar graphs are more susceptible to bias from y-axis distortions than dot or line graphs.

Conclusions:

• Y-axis distortions significantly influence the perception of data in visualizations.
• The proposed 'anchors' framework can explain these perceptual biases.
• Understanding these effects is crucial for accurate data interpretation and effective communication across various visualization types.