Immersive Analytics (IA) uses immersive virtual and augmented reality displays for data visualization and visual analytics. Designers rely on studies of how accurately people interpret data in different visualizations to make effective visualization choices. However, these studies focus on data analysis in traditional desktop environments. We lack empirical grounding for how to best visualize data in immersive environments. This study explores how people interpret data visualizations across different display types by measuring how quickly and accurately people conduct three analysis tasks over five visual channels: color, size, height, orientation, and depth. We identify key quantitative differences in performance and user behavior, indicating that stereo viewing resolves some of the challenges of visualizations in 3D space. We also find that while AR displays encourage increased navigation, they decrease performance with color-based visualizations. Our results provide guidelines on how to tailor visualizations to different displays in order to better leverage the affordances of IA modalities.

Visualizations often encode numeric data using sequential and diverging color ramps. Effective ramps use colors that are sufficiently discriminable, align well with the data, and are aesthetically pleasing. Designers rely on years of experience to create high-quality color ramps. However, it is challenging for novice visualization developers that lack this experience to craft effective ramps as most guidelines for constructing ramps are loosely defined qualitative heuristics that are often difficult to apply. Our goal is to enable visualization developers to readily create effective color encodings using a single seed color. We do this using an algorithmic approach that models designer practices by analyzing patterns in the structure of designer-crafted color ramps. We construct these models from a corpus of 222 expert-designed color ramps, and use the results to automatically generate ramps that mimic designer practices. We evaluate our approach through an empirical study comparing the outputs of our approach with designer-crafted color ramps. Our models produce ramps that support accurate and aesthetically pleasing visualizations at least as well as designer ramps and that outperform conventional mathematical approaches.

Scatterplots commonly use multiple visual channels to encode multivariate datasets. Such visualizations often use size, shape, and color as these dimensions are considered separable--dimensions represented by one channel do not significantly interfere with viewers' abilities to perceive data in another. However, recent work shows the size of marks significantly impacts color difference perceptions, leading to broader questions about the separability of these channels. In this paper, we present a series of crowdsourced experiments measuring how mark shape, size, and color influence data interpretation in multiclass scatterplots. Our results indicate that mark shape significantly influences color and size perception, and that separability among these channels functions asymmetrically: shape more strongly influences size and color perceptions in scatterplots than size and color influence shape. Models constructed from the resulting data can help designers anticipate viewer perceptions to build more effective visualizations.