Exploring Data Distributions: Visual Design And Evaluation

Visual overviews of tables of numerical and categorical data have been proposed for tables with a single value per cell. This article addresses the problem of exploring tables with columns that consist of cells that are distributions, for example, the distributions of movie ratings or trust ratings in recommender systems, age distributions in demographic data, usage distributions in logs of telephone calls, and so on. This article expands on heatmap approaches and proposes a novel way of displaying and interacting with distribution data. The usability study demonstrates the benefits of the heatmap interface in providing an overview of the data and facilitating the discovery of interesting clusters, patterns, outliers and relationships between columns.