A framework for value visualizations
Regular talk, 3:50 - 4:50 PM
This introduces a framework for value visualizations, displays where the placement of the plotted objects is designed to compare numeric values, Most visualization systems construct value visualizations with low-level instructions or a compositional grammar based on the shapes of the plotted objects. LessR package, however, organizes value visualizations using only 3 core functions, which reflect the structure of the underlying data rather than the underlying shapes, each answering a specific question: Chart() for aggregated numeric values across categories to illustrate how many or to what extent, X() for numeric distributions plotted along the x-axis, and XY() for numeric relationships plotted in the x-y coordinate system. Each function supports multiple analytically equivalent but perceptually distinct geometric forms, optionally stratified by grouping factors according to the parameter by for within-panel grouping and the parameter facet for between-panel grouping. Further, these functions share only those two parameters, plus x and perhaps y for the primary variables, and parameter type for specifying the specific geometric form, such as histogram, substantially reducing syntactic complexity. Three functions using at most five parameters provide a concise, consistent vocabulary across the functions for a large variety and types of data visualizations, many interactive, that range from density plots, sunburst charts, bivariate contour curves, time series, and many more.
 |
Pronouns: he/himPortland, OR, USAPh.D in what would now be called Data Science from Michigan State University in 1979. Currently, Professor in the Masters of Applied Data Science for Business (MSADSB) program at Portland State University, primarily responsible for the courses in machine learning, deep learning and neural networks, and data visualization. Have published a wide range of articles in the behavioral and data science journals. |