Graph schema and best graph type to compare discrete groups: Bar, line, and pie

Area of Science:

• Cognitive Psychology
• Human-Computer Interaction
• Data Visualization

Background:

• Graph schemas influence the effectiveness of data visualization for group comparisons.
• Understanding whether graph schemas are distinct or share common structures is crucial for optimizing data presentation.

Purpose of the Study:

• To investigate if graph schemas are based on perceptual features or common invariant structures.
• To determine the optimal graph type (bar, line, or pie) for comparing discrete groups.
• To examine the extent to which different graph types share common schemas.

Main Methods:

• A switching paradigm was employed across three experiments, comparing pairs of graph types (bar vs. line, bar vs. pie, line vs. pie).
• Participants performed group comparisons on data presented in graphs, and reaction times (RTs) were measured.
• Switch costs (prolonged RTs when switching graph types) were analyzed to infer differences in graph schemas.

Main Results:

• Switch costs were observed for all graph type pairings, indicating no complete sharing of schemas.
• Smaller switch costs between bar and line graphs suggest greater overlap in their schemas compared to other pairings.
• Bar graphs facilitated the fastest group comparisons, outperforming line and pie graphs.

Conclusions:

• Graph schemas appear to be hierarchical, comprising both shared and type-specific components.
• Bar graphs are the most suitable type for comparing discrete groups due to superior performance.
• While distinct schemas exist, there is partial overlap, particularly between bar and line graphs.