论文标题
混合响应的多元回归以评估可视化设计
Multivariate Regression of Mixed Responses for Evaluation of Visualization Designs
论文作者
论文摘要
信息可视化通过图形表示复杂的数据集可显着增强人类的感知。各种可视化设计使有效评估满足用户的偏好和特征的所有可能设计的各种挑战。大多数现有的评估方法进行了用户研究，以通过问卷和访谈从用户那里获得多元定性答复。但是，这些方法无法支持对设计的在线评估，因为它们通常很耗时。需要一个统计模型，以根据非干预测量（即可穿戴传感器信号）来预测用户对可视化设计的偏好。在这项工作中，我们提出了混合反应（MRMR）的多元回归，以促进可视化设计的定量评估。提出的MRMR方法能够提供有意义的变量选择的准确模型预测。进行了14个有效参与者的可视化设计的模拟研究和用户研究，以说明所提出的模型的优点。
Information visualization significantly enhances human perception by graphically representing complex data sets. The variety of visualization designs makes it challenging to efficiently evaluate all possible designs catering to users' preferences and characteristics. Most of existing evaluation methods perform user studies to obtain multivariate qualitative responses from users via questionnaires and interviews. However, these methods cannot support online evaluation of designs as they are often time-consuming. A statistical model is desired to predict users' preferences on visualization designs based on non-interference measurements (i.e., wearable sensor signals). In this work, we propose a multivariate regression of mixed responses (MRMR) to facilitate quantitative evaluation of visualization designs. The proposed MRMR method is able to provide accurate model prediction with meaningful variable selection. A simulation study and a user study of evaluating visualization designs with 14 effective participants are conducted to illustrate the merits of the proposed model.
