Visualization Analysis and Design provides a systematic, comprehensive framework for thinking about visualization in terms of principles and design choices. The book features a unified approach encompassing information visualization techniques for abstract data, scientific visualization techniques for spatial data, and visual analytics techniques for interweaving data transformation and analysis with interactive visual exploration. It emphasizes the careful validation of effectiveness and the consideration of function before form. The book is suitable for a broad set of readers, from beginners to more experienced visualization designers. It does not assume any previous experience in programming, mathematics, human-computer interaction, or graphic design and can be used in an introductory visualization course at the graduate or undergraduate level.

We contribute the first systematic approach to designing visualization snippets: compact previews of interactive visualization bundles, used by millions of people on platforms such as Tableau Online or Power BI. We present a computational pipeline to compress the visual and textual content of bundles into representative previews that is adaptive to a provided pixel budget and provides high information density with multiple images and carefully chosen keywords. We also reflect on the method of visual inspection through random sampling to gain confidence in model and parameter choices.

Short paper on the design and implementation of TimeElide, a visual analysis tool for the novel data abstraction of non-contiguous time series slices, namely time intervals that contain a sequence of time-value pairs but are not adjacent to each other.

Graph neural networks (GNNs) are a class of powerful machine learning tools that model node relations for making predictions of nodes or links. We propose an approach to corresponding an input graph to its node embedding (aka latent space), a common component of GNNs that is later used for prediction. We abstract the data and tasks, and develop an interactive multi-view interface called CorGIE to instantiate the abstraction. As the key function in CorGIE, we propose the K-hop graph layout to show topological neighbors in hops and their clustering structure.

We present an algorithm for data reconnaissance through automatic visualization recommendation, GEViTRec. Our approach handles a broad variety of dataset types and automatically generates coordinated combinations of charts, in contrast to existing systems that primarily focus on singleton visual encodings of tabular datasets. We automatically detect linkages across multiple input datasets by analyzing non-numeric attribute fields, creating an entity graph within which we analyze and rank paths. For each high-ranking path, we specify chart combinations with spatial and color alignments between shared fields, using a gradual binding approach to transform initial partial specifications of singleton charts to complete specifications that are aligned and oriented consistently. A novel aspect of our approach is its combination of domain-agnostic elements with domain-specific information that is captured through a domain-specific visualization prevalence design space.

Journal version of SepEx EuroVA workshop paper.

We conduct a technical observation study of 50 public repositories of data and analysis code authored by 33 professional journalists at 26 news organizations. We develop two detailed and cross-cutting taxonomies of data wrangling in computational journalism, for actions and for processes. We observe the extensive use of multiple tables, a notable gap in previous wrangling analyses. We develop a concise, actionable framework for general multi-table data wrangling that includes wrangling operations documented in our taxonomy that are without clear parallels in other work.

We discuss the challenges associated with creating a content-based recommendation system based on visualization specifications and explore more concretely how to implement the relevance measures required using Tableau workbook specifications as the source of content data. We also demonstrate what information can be extracted from these visualization specifications and how various natural language processing techniques can be used to compute similarity between workbooks as one way to measure relevance.

We introduce the notion of a data-first design study which is triggered by the acquisition of real-world data instead of specific stakeholder analysis questions.

We introduce low-stretch trees to the visualization community with LSQT, our novel technique that uses quasi-trees for both layout and edge bundling. Our method offers strong computational speed and complexity guarantees by leveraging the convenient properties of low-stretch trees, which accurately reflect the topological structure of arbitrary graphs with superior fidelity compared to arbitrary spanning trees. Low-stretch quasi-trees also have provable sparseness guarantees, providing algorithmic support for aggressive de-cluttering of hairball graphs. LSQT does not rely on previously computed vertex positions and computes bundles based on topological structure before any geometric layout occurs.

We present SepEx, an interactive visualization approach for the assessment and comparison of class separation measures for multiple datasets. SepEx supports analysts with the comparison of multiple separation measures over many high-dimensional datasets, the effect of dimensionality reduction on measure outputs by supporting nD to 2D comparison, and the comparison of the effect of different dimensionality reduction methods on measure outputs.

We present detailed data and task models that capture the demands of group decision making in particular, going through the four stages of the nested model of visualization design to develop and test a prototype to support group decision making when decision makers express their preferences directly on the alternatives.

We present a multi-year design study that resulted in Ocupado, a set of visual decision-support tools centered around occupancy data for stakeholders in facilities management and planning. Ocupado uses WiFi devices as a proxy for human presence, capturing location-based counts that preserve privacy without trajectories. We contribute data and task abstractions for studying space utilization for combinations of data granularities in both space and time. In addition, we contribute generalizable design choices for visualizing location-based counts relating to indoor environments.

We propose a multi-level segmentation model that works for many image editing tasks including poster creation, portrait retouching, and special effect creation. The lowest-level chunks of logged events are computed using a support vector machine model and higher-level chunks are built on top of these, at a level of granularity that can be customized for specific use cases. Our model takes into account features derived from four event attributes collected in realistically complex Photoshop sessions with expert users: command, timestamp, image content, and artwork layer.

We propose an area-aware approach to clutter metrics that tracks the extent of geometric overlaps between node-node, node-edge, and edge-edge pairs in detail. It handles variable-size nodes and explicitly treats metanodes and leaf nodes uniformly. We call the combination of a sprawl metric and an area-aware clutter metric a sprawlter metric.

We identify data reconnaissance and task wrangling as processes that experts undertake to discover and identify sources of data that could be valuable for some specific analysis goal.We propose a conceptual framework with a four-phase cycle of acquire, view, assess, and pursue that occurs within three distinct chronological stages, which we call fog and friction, informed data ideation, and demarcation of final data.

Segmentifier supports an iterative process of refining collections of action sequences into meaningful segments, to make sense of large and noisy real-world e-commerce clickstream datasets.

We present the new visual encoding idiom of aggregated dendrograms to concisely summarize the topological relationships between interactively chosen focal subtrees according to biologically meaningful criteria, and provide a layout algorithm that automatically adapts to the available screen space.

We present a method for systematically surveying data visualizations using the analysis of both text and images, and the result of applying this method to the domain of genomic epidemiology.

Given a PubMed-compatible search query, Adjutant downloads the relevant articles and allows the user to perform an unsupervised clustering analysis to identify data-driven topic clusters. Following clustering, users can also sample documents using different strategies to obtain a more manageable dataset for further analysis.

We discuss we learned across two projects involving spatiotemporal data that does not use the data abstraction of trajectories: bike-sharing networks and human occupancy data for buildings. We discuss the general implications of this data type in terms of ethics, data preprocessing, tasks, and visual encodings.

The Gaze Reading Self-Interruption Visualizer integrates a visualization front-end suitable for data cleansing and a prediction back-end that can be run repeatedly as the input data is iteratively improved.

The patch lineage graph encapsulates the history of cellular colonies by exploiting temporal and spatial coherence of subpopulations using a particle-based approach, for datasets where the traditional approach of single-cell detection fails due to high density, and high diversity of cell shapes and numbers, for high-resolution images with high amounts of noise.

Evidence-based guidelines for communicating tuberculosis genomic test results, using rigorous methods for generating and testing alternative formats based on iterative and human-centered visualization design study methodology.

Framework based on analysis reports derived from open-coding 20 design study papers published at IEEE InfoVis 2009-2015, to build on the previous work of abstractions that collectively encompass a broad variety of domains.

Discussion of adapting visualization methodology in the highly regulated fields such as public health, with a case study on tuberculosis clinical care through the BC Centre for Disease Control.

An exploration of the design space of timelines for presentation.

Redesign study in the energy analysis domain.

Authoring tool for journalists and others who want to quickly create timelines from free-form text.

Visual analysis of group cohesion in multiplex networks through dual linked views, featuring the new data abstractions of substrate and catalyst networks.

We characterize five task sequences related to visualizing dimensionally-reduced data, drawing from data collected from interviews with ten data analysts spanning six application domains, and from our understanding of the technique literature.

A predictive model based on extensive qualitative analysis from an observational study, with preliminary validation through a hierarchical multiple regression analysis comparing graph readability factors.

A design study about Overview, an application for the systematic analysis of large document collections based on document clustering, visualization, and tagging.

The QSNE document-focused dimensionality reduction algorithm combines stochastic neighborhood embedding with query techniques from information retrieval.

Journal version of BELIV 2012 workshop paper on nested model extension; additions are analysis of three papers from InfoVis12 with the NBGM, and an examination of design models in previous work.

Quantitative data study analyzing three types of scatterplots (2D, SPLOM, and 3D) and four types of DR choices. Key findings: 2D is often good enough, and switching DR techniques is the best next step; SPLOMs occasionally add value; interactive 3D rarely helps and often hurts.

Framework for analyzing individual and compound tasks in terms of why-what-how.

Design study for genomic data proposing alternative to standard genome browser approach.

An extension to the four-level Nested Model for design and validation of visualization systems that defines the term guidelines in terms of blocks at each level.

Glint is designed to automatically minimize the total number of distances computed by progressively computing a more and more densely sampled approximation of the distance matrix.

We provide definitions, propose a methodological framework, and provide practical guidance for conducting design studies.

Design study focused on supporting automotive engineers who need to specify and optimize traffic patterns for in-car communication networks. The task and data abstractions for actively making changes to an overlay network, where logical communication specifications must be mapped to an underlying physical network network, are very different than the dominant use cases from the domain of social network analysis.

Qualitative evaluation of 800+ plots, including analysis of the reasons for failure of previous cluster separation metrics and a taxonomy of factors that affect separation.

Design study supporting the analysis of fisheries simulation data, including sensitivity analysis, global trade-offs analysis, and staged uncertainty.

The proposed design space is spanned by two axes: recall support, and absolute rating vs. relative ranking.

Invited paper discussing Cerebral and Pathline design rationale in detail in the context of the Nested Model framework.

Reflections on the design of a system for visualizing a high-dimensional environmental sustainability dataset. Despite some successes on designing visual encoding and interaction techniques, QuestVis was ultimately not deployed because of a mismatch between the design goals of the project and the true needs of the target user community. We analyze this breakdown in the context of the Nested Model framework, which was motivated in part by this experience.

This monograph provides design guidelines for designing multi-level interfaces based on a fine-grained analysis of of 22 empirical studies. We considered three multi-level interface types in this synthesis review: temporal, or temporal switching of the different levels as in zooming interfaces; separate, or displaying the different levels simultaneously but in separate windows as in overview + detail interfaces; and embedded, or showing the different levels in a unified view as in focus + context interfaces.

Design study about a visualization tool that supports exploration of multiple possible computational summaries that mix spatial information about each cell in a developing fruit fly embryo, gene expression measurements over time, and data from multiple related species or organisms.

Dimensionality reduction for the rest of us: system paper about toolkit for dimensionality reduction providing local and global guidance to users who may not be experts in the mathematics of high-dimensional data analysis.

Journal version of PacificVis TugGraph paper.

Design study of a visualization tool for comparative functional genomics supporting simultaneous analysis of functional, pathway, and phylogenetic data. We present two new visual encoding techniques: linearized metabolic pathways, and curvemaps.

This model provides a unified framework for considering both the design and the validation of visualization systems at four cascading levels, with prescriptive advice on determining appropriate evaluation approaches by identifying threats to validity unique to each level.

Design study on the creation of a multiscale synteny browser for exploring conservation relationship in comparative genomics data.

TugGraph is a system for exploring paths and proximity around nodes and subgraphs in a multilevel graph.

A book chapter for an undergraduate computer graphics textbook summarizing process, principles, and techniques for visualization with an emphasis on abstract data.

Glimmer is a new multilevel multidimensional scaling algorithm that exploits the GPU. We demonstrate its benefits in a detailed comparison against several previous algorithms.

Paper for the biological community describing InnateDB, including the Cerebral visualization component.

We describe the data information display needs of immunologists and describe the design decisions used to create Cerebral, a system that incorporates experimental data directly into the graph display, using the biologically guided graph layout announced in the previous appnote.

GrouseFlocks supports interactive exploration of the space of possible hierarchies for an input graph with domain-specific attributes at the nodes. The system generates hierarchies that reflect the underlying graph topology by requiring that subgraphs respect edge and connectivity conservation.

Based on our experience in extracting design guidelines from existing quantitative studies, we recommend improvements to both study design and reporting to promote meta-analysis.

The LiveRAC visualization system supports the analysis of large collections of time-series data with hundreds of parameters across thousands of network devices. It provides high information density using a reorderable matrix of charts, with semantic zooming adapting each chart's visual representation to the available space.

A book chapter exhorting infovis authors to avoid pitfalls at several stages of research process, including visual encoding during design, a checkpoint before starting to write, and after a full paper draft exists. The paper page includes my current list of favorite design study examples.

For the task we studied, point-based spatializations were far superior to landscapes, and 2D landscapes were superior to 3D landscapes.

We compared four interfaces: overview-only, detail-only, separate overview and detail windows, and detail embedded within overview as in Focus+Context interfaces.

Taking a multiple-coordinated view approach, Session Viewer shows multiple session populations at the Aggregate, Multiple, and Detail data levels to support different analysis styles.

Interactive and steerable exploration of multilevel graph hierarchies, as created by TopoLayout.

Short note aimed at the bioinformatics community announcing a graph layout approach for protein-protein interaction networks that exploits annotations about where genes/proteins are expressed within a cell and their function.

Multilevel graph drawing algorithm where the original graph is decomposed into features based on topological structures such as trees and clusters, each feature is drawn with an appropriately tuned algorithm, and passes for crossing reduction and overlap elimination reduce visual clutter.

Two-level approach to drawing complex graphs where an area-aware version of the RINGS tree drawing algorithm is used to show the high-level structure, and the force-directed LGL approach is used to show the low-level structure.

Navigation for accordion drawing, whereas previous papers focused on rendering.

Evaluation of how both linear and nonlinear geometric transformations affect visual memory, as opposed to previous paper on visual search.

Evaluation comparing navigation techniques and availability of contextual information.

Five-page summary of VRC report.

Five-page summary of VRC report.

This report, sponsored by the NIH and NSF, is a followup to the 1987 NSF Visualization report. Our goal is to evaluate the progress of the maturing field of visualization, to help focus and direct future research projects, and to provide guidance on how to apportion national resources. Our findings and recommendations reflect not only information gathered from visualization and applications scientists during two workshops on Visualization Research Challenges but also input from the larger visualization community.

Longer journal version of PRISAD paper, recommended over the conference version. The algorithm explanation is clearer and more detailed.

Conference version of PRISAD paper, presenting generic and efficient infrastructure for accordion drawing with an emphasis on rendering.

We quantify the perceptual cost of geometric transformations for visual search tasks, finding no-cost and low-cost zones for fisheye transformations.

Visual diff for trees, introducing accordion drawing.

In this short introduction I briefly describe the goals of the field of information visualization to set the stage for the four accepted papers in this special issue, for which I was the guest editor.

This paper discusses a user study conducted at Microsoft Research that found a statistically significant improvement in task time when a novel web browser that included the H3Viewer was compared to more traditional browsers.

My dissertation includes a chapter each on the H3, Planet Multicast, and Constellation projects, with much more analysis than appeared in any of the original papers.

A paper on interactive rendering of scenes where objects can have different projections, as in the simulaneous multiple viewpoints of Cubist paintings.

A paper on a 2D interactive system that uses a custom graph layout algorithm and many perceptual channels to help computational linguists debug their algorithms for generating and using large semantic networks.

A paper which presents the H3Viewer guaranteed frame rate drawing algorithm, along with a brief review of the H3 layout algorithm for context. Very large graphs can be navigated at a constant frame rate, as long as the entire graph can fit into main memory. The viewer is custom OpenGL/C++ code. My implementation of the algorithm discussed here was shipped with the Site Manager 1.1 software from SGI. The source is available for free noncommercial use.

An article which briefly reviews the H3 layout and H3Viewer drawing algorithms. The main new material here is a discussion about possible tasks for graph drawing beyond the global overview problem.

A paper that presents a much improved layout algorithm that exploits some of the properties of 3D hyperbolic space to achieve reasonable information density. The viewer is a highly modified version of Geomview, which supports manual subtree collapse and expansion. My implementation of the algorithm discussed here was shipped with the Site Manager 1.0 software from SGI.

A case study describing the Planet Multicast project: an interactive 3D geographic visualization of the Internet's multicast backbone, where tunnels are shown as arcs on a globe.

The first of my hyperbolic papers. The layout algorithm is the the most straightforward extension of cone trees to 3D hyperbolic space. The viewer is Geomview/WebOOGL. The system allows neither manual nor automatic collapsing of the graph, so scalability is extremely limited.

A survey article about interactive mathematical visualization, includes many pictures.