Glyph Design for Temporal and Multi-Dimensional Data : Design Considerations and Evaluation

Abstract

The goal of this thesis is to provide researchers and practitioners with guidance in designing data glyphs for temporal and multi-dimensional data. Therefore, the term "glyph'' in the context of information visualization has to be introduced and defined first, to establish a common understanding of the overall topic and motivate the need for additional support in selecting or creating data glyphs. This definition is the basis for reviewing literature about data glyph experiments, for conducting further controlled user studies, and finally for introducing new data glyph designs. In summary, the computer science contributions in the area of information visualization are threefold.First, literature about quantitative experiments on data glyphs from the past 70 years is systematically reviewed. By sampling and tabulating the literature on data glyph studies, listing their designs, questions, data, and tasks an overview about study goals and results is provided and open research gaps are revealed. Based on this meta analysis of all results a catalog of design considerations is created, which will be further extended throughout this thesis.Second, the previously identified research gaps are used as a motivation for conducting controlled user studies, which are introduced in this thesis. Since variations of star glyphs and radial color encodings have not received much research attention, these designs will be subject to quantitative experiments. Results indicate that, against intuition, the whisker glyph which is hardly used in practice outperforms the alternative star glyph variations. Additionally, further study outcomes suggest that radial glyph layouts making use of the visual variable orientation to separate different dimensions are the best choice for detecting specific points in time. This finding contradicts the ranking of visual variables from Cleveland and McGill where position encodings outperform orientation encodings. Based on these results the set of design considerations collected in the initial survey is extended and summarized to facilitate the guidance in creating and selecting data glyph designs.Third, the design space of data glyphs is enriched with two new metaphoric designs tailored towards specific domains and evaluated with use cases and controlled user studies to show their applicability to real-world scenarios. The clock glyph representation, for example, supports the analyst in detecting specific points in time by arranging the temporal dimensions in a radial fashion. Results from quantitative experiments indicate the usefulness of this metaphoric approach outperforming well-established alternative representations like line glyphs. The leaf glyph technique on the other hand makes use of environmental cues to encode multi-dimensional data controlling main leaf properties like leaf morphology, leaf venation, and leaf boundary. The design is motivated by the human ability to visually discriminate natural shapes like trees in a forest, single flowers in a flowerbed, or leaves at shrubs. Due to its aesthetically pleasing appearance, this design is suitable for being used in mass media and data journalism. A use case scenario with forest fire data reveals the strengths of this design being effectively interpretable for storytelling in environmental data analysis.