Dataset for Evaluating Glyph Design Sampled from Quantum Spins Simulations

This dataport will be useful to those interested in visual design for complex physics phenomena. We have included two quantum physics data sample datasets and our empirical study results. (1) evaluation results from two experiments to empirically validate that separable bivariate pairs of large-magnitude-range vector magnitude representations are more efficient than integral pairs. (2) quantum physics simulation data used in the first experiment with twenty participants. The simulation data contains more than 10K vector spins, and the samples contain about 445-550 spins. In the study, participants performed three local tasks requiring reading no more than two glyphs. These quantum spins are large in magnitude range and thus were shown to participants (physicists) as a series of separable bivariate pairs. These spin vector studies supported that separable-pair visualization led to the most accurate answers and the shortest task execution time, while integral ones were among the least accurate;   unless a redundant categorical color feature was added.} (3) quantum simulation data from the second study when we scale up the search space and when participants must look at the entire scene of hundreds of vectors to get an answer.We measured if more separable bivariate pairs can help viewers obtain scene structures first. Eighteen participants used three separable pairs in three global tasks: find a specific target in $20$ seconds, locate the maximum magnitude in 20 seconds, and estimate the total number of regions of interests (here vector exponents) within 2 seconds. Our results revealed that the higher the separability, the higher the accuracy. We believe that the reason was that the separable glyph pairs introduced emergent global (and preattentive) scene features for viewers to manage complexity.Associated source code, training documents, participants' results collected, statistical methods, and results at https://osf.io/4xcf5/?view_only=94123139df9c4ac984a1e0df811cd580}{$https://osf.io/4xcf5/?view_only=94123139df9c4ac984a1e0df811cd580$ for reproducible research.

本数据集资源库面向关注复杂物理现象可视化设计的研究者与从业者，具备较高的参考价值。本资源库内含两套量子物理数据样本数据集与配套实证研究成果：
(1) 两项实验的评估结果，用于实证验证：当采用大模长范围（vector magnitude）向量的可分离二元对（separable bivariate pairs）表示时，其效率优于整体系表示（integral pairs）。
(2) 第一项实验所用的量子物理模拟数据，共有20名受试者参与。该模拟数据包含超10000个向量自旋样本，单组样本包含约445至550个自旋量。本研究中，受试者需完成三项局部任务，每项任务仅需读取不超过两个图形标识（glyph）。由于这些量子自旋的模长分布范围较大，因此以可分离二元对的形式展示给受试者（物理学家群体）。该自旋向量研究证实，可分离二元对可视化方案可带来最高的作答准确率与最短的任务执行时长，而整体系可视化方案的准确率则处于较低水平——除非额外添加冗余的分类颜色特征。
(3) 第二项研究所用的量子模拟数据，该研究扩大了搜索空间，要求受试者需观察包含数百个向量的完整场景以获取答案。本次研究旨在验证更多可分离二元对是否可帮助观者优先识别场景结构。共有18名受试者参与三项全局任务：在20秒内找到指定目标、在20秒内定位模长最大值，以及在2秒内估算感兴趣区域（regions of interests，此处指代向量指数）的总数量。研究结果显示，可分离性越高，准确率也越高。我们认为其原因在于，可分离图形标识对为观者引入了涌现式全局（及前注意preattentive）场景特征，从而帮助其处理复杂度问题。
相关源代码、训练文档、收集的受试者实验结果、统计方法与全部研究成果均可通过以下链接获取：https://osf.io/4xcf5/?view_only=94123139df9c4ac984a1e0df811cd580，以供可重复研究使用。