To Cut or Not To Cut? A Systematic Exploration of Y-Axis Truncation

Title of the Paper

To Cut or Not To Cut? A Systematic Exploration of Y-axis Truncation

Paper Information

• Domain: Data Visualization, Graphical Design
• Keywords: Y-axis truncation, deceptive visualization, structure-preserving transformation, algebraic visualization design, bar charts, user perception

Research Background and Problem

• Problem and Challenges: Y-axis truncation is a controversial graphical practice. Some studies suggest that this method misleads viewers' subjective perception of trends or magnitude changes, but its impact on users' task accuracy has not been systematically studied.
• Significance: Graphical design and data visualization play critical roles in scientific communication and decision support. Understanding the contexts in which Y-axis truncation is appropriate can help design visualization tools that convey information effectively while avoiding misleading representations.
• Motivation: Most existing research on Y-axis truncation focuses on subjective perception (e.g., trend magnitude) rather than its impact on users' ability to perform specific tasks. This study combines visualization theory with systematic theoretical analysis and empirical evaluation of Y-axis truncation.

Solution

• Main Methods:
  • Define the "monotonicity" of Y-axis truncation: whether truncation allows data relationships to remain visually consistent. Monotonic truncation preserves the original data relationships, while non-monotonic truncation may distort them.
  • Use Algebraic Visualization Design (AVD) to analyze the structural correspondence between data, tasks, and visualization forms.
  • Design experiments to evaluate user task performance under monotonic and non-monotonic truncation.
• Innovations:
  • Introduce the concept of structure-preserving transformations, establishing a mathematical link between monotonic truncation and data relationships.
  • Employ Bayesian multilevel regression models to assess the specific effects of truncation on user perception and task performance.
• Implementation Steps and Key Techniques:
  1. Classify Y-axis truncation based on monotonicity and data relationships.
  2. Design four judgment tasks (gap comparison, ratio comparison, gap estimation, ratio estimation) to reflect common user needs.
  3. Generate visual stimuli using different data distributions (Beta distribution and truncated Pareto distribution) to simulate real-world scenarios.
  4. Recruit participants for online experiments and collect task performance data.
  5. Analyze results using Bayesian statistical methods.

Research Findings

• Key Results:
  • Gap Tasks (e.g., comparing or estimating the height difference between bars):
    • Truncation has minimal impact on user performance, regardless of monotonicity.
    • Non-monotonic truncation causes slight bias in gap estimation, but the effect is minor.
  • Ratio Tasks (e.g., comparing or estimating percentage changes):
    • Monotonic truncation performs on par with no truncation, and both significantly outperform non-monotonic truncation.
    • Non-monotonic truncation significantly reduces user accuracy.
• Comparison with Existing Research:
  • Partially validates prior findings on Y-axis truncation's effects on subjective perception, but the distinction between gap and ratio tasks adds depth to this field.
• Limitations:
  • Did not investigate the cognitive processes underlying perception.
  • Did not examine the effects of visual framing or subtle cues on results.
  • Task scope was limited to numerical perception, excluding subjective trend perception or time-constrained conditions.
• Future Directions:
  • Validate findings in a broader range of task contexts.
  • Further explore the relationship between truncation and subjective perception.
  • Investigate interactive visualization designs for Y-axis truncation.

Discussion and Recommendations

• Design Recommendations:
  • For tasks that are well-defined and do not involve ratio changes, monotonic truncation can optimize visual design or focus perception.
  • For tasks involving ratio changes, to avoid visual misrepresentation, use monotonic truncation or no truncation.
  • When task purposes are unclear, avoid fixed truncation and adopt interactive designs to balance flexibility and information transparency.
• Theoretical Contributions:
  • Proposes a new theoretical framework for analyzing Y-axis truncation (D ↔ V ↔ T model).
  • Provides quantitative evaluation of the impact of monotonic truncation on user performance, complementing prior studies focused solely on subjective perception.
• Practical Implications:
  • Offers clear guidance for chart designers based on data relationships and user task requirements.
  • Advances research on the interaction between "subjective perception" and "objective judgment" in data visualization.