Missing data pose a widely known challenge for multivariate data analysis. Imputation is a mechanism for replacing missing values with plausible substitutes. Multiple imputation is often used to generate several plausible datasets for a wide range of statistical models to create robust standard errors and reduce potential bias from complete case analyses. Implementing an imputation algorithm requires several decisions about settings to generate meaningful imputed datasets. The analysis tasks can be summarized as analyzing attribute relations, analyzing the missingness mechanisms, supporting the selection of imputation parameters, and analyzing the imputation outcome. We propose a visual analytics pipeline that supports decision-making in each of the required steps. Each step is supported by novel visual encodings or existing ones that are adapted to the task. In particular, we adapt all methods to operate on heterogeneous data. Key visualizations are a 2D embedding of attribute similarities for heterogeneous data, a heatmap visualization providing a scalable overview of the effect of missingness on data distributions between all pairs of attributes (and between all original and all imputed attributes), and extended parallel coordinates that can handle both heterogeneous and missing data. Moreover, we propose a novel scheme for computing the imputation order based on an interactive predictor graph visualization. We illustrate the effectiveness of our visual encodings and the interactive visual analytics pipeline within workflows executed with domain experts from epidemiology to analyze cohort studies. The domain experts reported increased confidence in the imputation result due to improved transparency.

The application of parallel axes for the interactive visual analysis of multidimensional data is a widely used concept. While multidimensional data sets are commonly heterogeneous in nature, i.e. data items contain both numerical and categorical (including ordinal) attribute values, the use of parallel axes often assumes either numerical or categorical attributes. While Parallel Coordinates and their large variety of extensions focus on numerical data, Parallel Sets and related methods focus on categorical attributes. While both concepts allow for displaying heterogeneous data, no clear strategies have been defined for representing categories in Parallel Coordinates or discretization of continuous ranges in Parallel Sets. In practice, type conversion as a pre-processing step can be used as well as coordinated views of numerical and categorical data visualizations. We evaluate traditional and state-of-the-art approaches with respect to the interplay of categorical and numerical dimensions for querying probability-based events. We also compare against a heterogeneous Parallel Coordinates/Parallel Set approach with a novel interface between categorical and numerical axes . We show that approaches for mapping categorical data to numerical axis representations can lead to lower accuracy in answering probability-based questions and higher response times than hybrid approaches in multiple-event scenarios.

While different automated procedures for vowel formant prediction have recently been proposed, it is unclear how reliably these methods perform in the phonetic study of vowels in New Englishes and how such approaches could be applied to specific varieties. This paper compares different automatic methods for vowel formant prediction in New Englishes, using manual measurements of Trinidadian English as a baseline. The results show that all methods perform significantly better than default formant parameters often used in speech analysis packages, and that a Bayesian formant tracker calibrated with American (US-FAVE) and Trinidadian English (TRINI-FAVE) generally provides better results than an automatic procedure that optimizes formant ceilings on a vowel- and speaker-specific level. TRINI-FAVE measures vowels characteristic of Trinidadian English most accurately. Phonetic studies of vowels in New Englishes can benefit from these methods.

Mixed data sets containing numerical and categorical attributes are nowadays ubiquitous. Converting them to one attribute type may lead to a loss of information. We present an approach for handling numerical and categorical attributes in a holistic view. For data sets with many attributes, dimensionality reduction (DR) methods can help to generate visual representations involving all attributes. While automatic DR for mixed data sets is possible using weighted combinations, the impact of each attribute on the resulting projection is difficult to measure. Interactive support allows the user to understand the impact of data dimensions in the formation of patterns. Star Coordinates is a well-known interactive linear DR technique for multi-dimensional numerical data sets. We propose to extend Star Coordinates and its initial configuration schemes to mixed data sets. In conjunction with analysing numerical attributes, our extension allows for exploring the impact of categorical dimensions and individual categories on the structure of the entire data set. The main challenge when interacting with Star Coordinates is typically to find a good configuration of the attribute axes. We propose a guided mixed data analysis based on maximizing projection quality measures by the use of recommended transformations, named hints, in order to find a proper configuration of the attribute axes.

Principal curves are a long-standing and well-known method for summarizing large scatterplots. They are defined as self-consistent curves (or curve sets in the more general case) that locally pass through the middle of the scatterplot data. However, computing principal curves that capture well complex scatterplot topologies and are robust to noise is hard and/or slow for large scatterplots. We present a fast and robust approach for computing principal graphs (a generalization of principal curves for more complex topologies) inspired by the similarity to medial descriptors (curves locally centered in a shape). Compared to state-of-the-art methods for computing principal graphs, we outperform these in terms of computational scalability and robustness to noise and resolution. We also demonstrate the advantages of our method over other scatterplot summarization approaches.

Investigating etiology of a disease depends on the combination of tacit medical knowledge and multivariate analysis on a wide array of collected data. Confounding variables may generate a bias when exploring disease determinants, thus, reducing the predictive capabilities of risk factors. Stratified analysis is widely used in epidemiological settings to reduce the effect of confounding factors. We propose a stratified visual analysis approach based on linear projections and interactions in a Star Coordinates Plot (SCP), where the segregation power of dimensions in multiple strata can be explored interactively. We apply our approach to gain insight into three epidemiological results using stratified analysis regarding the prevalence of sleep apnea within age and gender strata and the segregating power of well-defined epidemiological risk factors.

Recently, large population-based studies gain increasing focus in the research community. Epidemiological studies acquire numerous data by means of questionnaires and examinations. Many of these studies also collect imaging data, for instance, magnetic resonance imaging or ultrasonography from hundreds or even thousands of participants. Here, we consider several on-going epidemiological studies conducted in Europe as well as challenges of subsequent image analysis and visualization of heterogeneous data, which were obtained within these studies. In particular, the main focus is on airway extraction tasks and the visual analytics problems. Available solutions and future directions for computer science specialists are presented and analyzed in terms of user-friendliness, speed, and efficiency.

Stenosis refers to the thinning of the inner surface (lumen) of vascular structures. Detecting stenoses and correctly estimating their degree is crucial in clinical settings for proper treatment planning. Such a planning involves a visual assessment, which in case of vascular structures is frequently based on 3D visual representations of the vessels. However, since vessel segmentation is affected by various sources of errors and noise in the imaging and image processing pipeline, it is crucial to capture and visually convey the uncertainty in a 3D visual representation. Moreover, it is crucial to quantify how much this uncertainty affects the calculated stenotic degree, since different severities lead to different treatments. We propose a novel approach for visualizing the shape deviation of different probability levels in vascular data, where the probability levels are computed from a probabilistic segmentation approach. Our non obstructive visual encoding is based on rendering a single opaque surface representing a probability level of the cumulative distribution function around the vessels' centerline. The surface rendering is enhanced with cumulative information about other levels. To do so, we traverse the probability space by applying an iterative projection method both inwards and outward until we reach surface variability within a given margin. We capture the shape variability between the different probability levels using the lengths of the projection lines, the change in angular directions, and the distortion of a parametrization. They are visually encoded using color and texture mapping. Furthermore, we allow for an interactive selection of a region of interest that automatically calculates the stenotic degree and how much the uncertainty affects the most likely result. We analyze our approach in comparison to state-of-the-art methods with medical experts in a study using both real magnetic resonance (MR) and computed tomography (CT) angiography data of vertebral arteries with stenoses as well as on MR angiography data with synthetically added stenoses and stenotic uncertainties. We evaluate how well our approach can guide medical experts in their assessment of the uncertainty in vertebral stenoses. (C) 2017 Elsevier Ltd. All rights reserved.

Scatterplot matrices (SPLOMs) are widely used for exploring multidimensional data. Scatterplot diagnostics (scagnostics) approaches measure characteristics of scatterplots to automatically find potentially interesting plots, thereby making SPLOMs more scalable with the dimension count. While statistical measures such as regression lines can capture orientation, and graph-theoretic scagnostics measures can capture shape, there is no scatterplot characterization measure that uses both descriptors. Based on well-known results in shape analysis, we propose a scagnostics approach that captures both scatterplot shape and orientation using skeletons (or medial axes). Our representation can handle complex spatial distributions, helps discovery of principal trends in a multiscale way, scales visually well with the number of samples, is robust to noise, and is automatic and fast to compute. We define skeleton-based similarity metrics for the visual exploration and analysis of SPLOMs. We perform a user study to measure the human perception of scatterplot similarity and compare the outcome to our results as well as to graph-based scagnostics and other visual quality metrics. Our skeleton-based metrics outperform previously defined measures both in terms of closeness to perceptually-based similarity and computation time efficiency.

To develop a search engine tailored to a specific country, its relevant Web resources have to be identified. The first step in this resource discovery is to retrieve Web documents that contain content related to the country. In our work, we aim towards a search engine for Honduras, a developing country in Latin America with an emerging use of the Web. In a preparatory step, we aim to find relevant domains prior to setting up our own crawler so that they can serve as seeds to the crawler, which can then be run in a more focused way to retrieve the actual Web pages. We initially use two angles for domain resource discovery. The first is to use the data available from the DMOZ catalogue, the second is to use major commercial search engines for an overview of Honduran domain names used on the Web. We report on initial results as well as the used methods and additionally give data about the geographic distribution of Honduran Web servers both inside and outside the country.