Better sensitivity to linear and nonlinear trends with position than with color

Area of Science:

• Data Visualization
• Human-Computer Interaction
• Cognitive Psychology

Background:

• Effective data visualization leverages human visual processing for information summarization.
• Previous design recommendations for pairwise comparisons may not generalize to trend detection tasks.

Purpose of the Study:

• To investigate whether design principles for pairwise comparisons extend to trend detection.
• To compare the efficacy of line graphs versus stripplots for identifying temperature trends over time.

Main Methods:

• Generated line graphs and stripplots from identical simulated temperature datasets with linear and exponential trends.
• Human observers performed a trend detection task, judging the direction (increasing/decreasing) of temperature changes.
• Assessed observer sensitivity and response bias for each graph type.

Main Results:

• Participants showed higher sensitivity to trend direction using line graphs compared to stripplots.
• A bias towards identifying increasing trends was observed with line graphs, decreasing as sensitivity increased.
• Despite superior performance with line graphs, stripplots were preferred aesthetically by over half of the participants.

Conclusions:

• Position-based encoding (line graphs) enhances trend detection performance over color-based encoding (stripplots).
• Aesthetic preferences for stripplots indicate a potential conflict between visual appeal and precise information communication.
• Future data visualization design should consider the balance between perceptual performance and user preference.