# Evaluating Alternative Glyph Design for Showing Large-Magnitude-Range Quantum Spins

**Authors:** Henan Zhao, Garnett W. Bryant, Wesley Griffin, Judith E. Terrill, Jian Chen  
**arXiv:** [2301.00002](https://arxiv.org/abs/2301.00002)  
**Date:** 2022-12-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2301.00002v1/

## Key points

- Experimental results on the use of bivariate glyphs for representing large-magnitude-range vectors
- Two key characteristics of the glyphs: separable visual dimensions and categorical representation
- Evaluation of how these conditions affect the effectiveness of the glyphs
- Two experiments conducted to assess performance: local tasks and global tasks
- Separable visual dimensions necessary for local tasks, but not sufficient for understanding large amounts of data
- Categorical representation crucial for perceiving global structures and examining complex datasets
- Categorical features in glyphs triggered optimal viewing behaviors according to participants' comments
- Addition of feather-like fishbone legends and tick-mark bands to enhance visualization with visual variable length
- Formulation of hypotheses suggesting lengthycolor feature-pair leads to more accurate answers, greater speed and accuracy compared to other separable dimensions, and redundancy in lengthycolor/lengthx pair reduces task completion time
- Three types of tasks performed by participants: magnitude reading, ratio estimation, and comparison
- Use of random sampling with modifications to replicate quantum physics data characteristics
- Within-subject design with two independent variables: feature-pair and task type employed for statistical analyses
- Contribution to development of perceptually accurate glyph representations for revealing patterns in large scientific datasets

## Summary

In this study, the researchers present experimental results on the use of bivariate glyphs for representing large-magnitude-range vectors. The glyphs have two key characteristics: (1) two visual dimensions that are separable; and (2) one of the visual dimensions uses a categorical representation, such as a categorical colormap. The aim of the study is to evaluate how these two conditions affect the effectiveness of the bivariate glyphs. The researchers conducted two experiments to assess the performance of the bivariate glyphs. In the first experiment, participants were asked to perform three local tasks that involved reading no more than two glyphs. The second experiment involved global tasks where participants had to examine hundreds of vector glyphs in order to find answers. The results of the experiments indicate that while separable visual dimensions are necessary for local tasks when comparing a few items, they are not sufficient for understanding large amounts of data. On the other hand, using a categorical representation is crucial for perceiving global structures and examining complex datasets. Participants' comments also revealed that the categorical features in the bivariate glyphs triggered optimal viewing behaviors. To enhance visualization when using visual variable length, feather-like fishbone legends and tick-mark bands around each cylinder were added by researchers. Based on their analysis and previous literature recommendations, several hypotheses were formulated by researchers which suggest that lengthycolor feature-pair can lead to more accurate answers, lengthycolor may lead to greater speed and accuracy compared to other separable dimensions, and redundancy in lengthycolor/lengthx pair can reduce task completion time. Participants performed three types of tasks: magnitude reading, ratio estimation, and comparison which required them to accurately determine magnitudes or ratios of vectors at specific points. Random sampling was used with modifications made to suit evaluation purposes while preserving domain-specific attributes in order to replicate quantum physics data characteristics; however modified data represented less than 4% of overall data points in each scene. A within-subject design with two independent variables: feature-pair and task type was employed by researchers for statistical analyses which evaluated performance of bivariate glyphs. Overall, this study contributes to development of perceptually accurate glyph representations for revealing patterns in large scientific datasets; source code, quantum physics data, training documents, participants' answers along with statistical analyses have been made available by researchers for reproducible science purposes.