Lost in Magnitudes: Exploring Visualization Designs for Large Value Ranges

Research Background and Problem

• Identified Issues or Challenges: In visualizing Order of Magnitude Values (OMVs), linear and logarithmic scale charts often fail to effectively represent data. Specifically, linear scales make smaller magnitude values indistinguishable, while logarithmic scales, though mitigating this issue, are not intuitive for the general public and scientists and are prone to misinterpretation. Furthermore, OMVs visualization designs inspired by scientific notation have not been widely explored.
• Significance: OMVs are widely applied in critical fields such as financial analysis, environmental monitoring, and pandemic data tracking. Improving the static visualization design of OMVs can more efficiently support data analysis, value comparison, and trend identification while making complex information more accessible to non-expert audiences.
• Research Motivation and Related Work: Existing OMVs visualization methods (e.g., Scale-Stacked Bar Chart, Order of Magnitude Markers) have made some improvements but still face issues, such as continuity problems between the mantissa and exponent components. Therefore, there is a need to systematically explore such visualization designs and establish design guidelines.

Solution

• Proposed Methods or Solutions:

  • A design space based on scientific notation was proposed, dividing OMVs into mantissa and exponent components and encoding them separately.
  • Four design guidelines were developed to enhance OMVs visualization effectiveness.
  • Two improved visualization designs (Facet and EplusM) were proposed and quantitatively compared with existing visualization methods.

• Innovations:

  • Systematically defined a constrained design space, generating all possible OMVs visualization designs based on the principles of the "Grammar of Graphics."
  • Introduced the "EplusM" scale, integrating mantissa and exponent into the same positional channel to achieve smooth numerical continuity.
  • Validated the feasibility and efficiency of the design space and formulated guidelines through a combination of qualitative and quantitative analyses.

• Implementation Steps:

  1. Design Space Modeling: Decomposed OMVs into mantissa and exponent based on scientific notation and combined them with other data attributes (nominal, ordinal, temporal, or quantitative) to generate the design space.
  2. Visualization Generation: Established generation rules to produce 336 eligible visualization designs.
  3. Qualitative Evaluation: Assessed the quality of visualization designs through an encoding-decoding process with an expert panel to identify potential issues.
  4. Quantitative Validation: Conducted crowdsourced experiments to compare the two new designs (Facet and EplusM) with other methods, analyzing their error rates, response times, and user confidence.

Research Outcomes

• Specific Results:

  • Systematically defined a design space for OMVs and generated applicable design guidelines through qualitative evaluation.
  • Proposed two designs, Facet (exponent encoding based on row facets) and EplusM (single-channel positional encoding of mantissa and exponent), and validated their superiority over existing methods.
  • Developed an open-source tool to facilitate the generation and evaluation of OMVs visualization designs for other researchers.

• Advantages Compared to Existing Solutions:

  • Facet and EplusM outperform logarithmic charts and Scale-Stacked Bar Charts in terms of numerical continuity between mantissa and exponent and higher encoding resolution in quantitative comparison tasks.
  • The study found that Facet and EplusM users exhibited higher confidence and reduced confusion between tasks.

• Experimental or Evaluation Results:

  • Accuracy: EplusM and Facet demonstrated significantly lower error rates in multiple tasks compared to logarithmic scale charts and Scale-Stacked Bar Charts.
  • Response Time: The response times for the two new designs were faster than those of the Scale-Stacked Bar Chart, particularly in numerical comparison tasks.
  • User Confidence: EplusM and Facet increased user confidence in completing tasks, reflecting their usability in OMVs visualization tasks.

• Limitations and Future Directions:

  • Limitations: The study covered only a small range of the design space and did not include interactive features, redundant encoding, or designs with multiple marker types. Additionally, evaluation tasks focused primarily on low-level perceptual tasks and did not address more complex data interaction or decision-making scenarios.
  • Future Directions: Expanding the scope of research to include more data types, more complex tasks, and studying the applicability of OMVs visualizations in real-world scenarios. Further exploration of enhancing aesthetics, trust, and user engagement in designs is also recommended.