Title: High Dimensional Brushing for Interactive Exploration of Multivariate Data

Authors: Allen R. Martin, Matthew O. Ward

Abstract: Brushing is an operation found in many data visualization systems. It is a mechanism for interactively selecting subsets of the data so that they may be highlighted, deleted, or masked. Traditionally, brushes have been defined in screen space via methods such as painting and rubberband rectangles. In this paper we describe the design of N-dimensional brushes which are defined in data space rather than screen space, and show how they have been integrated into XmdvTool, a visualization package for displaying multivariate data. Depending on the data display technique in use, brushes may be specified and manipulated via direct or indirect methods, and the specification may be demand-driven or data-driven. Various brush operations such as highlighting, linking, masking, moving average, and quantitative display have been developed to apply to the selected data. In addition, we have explored several new brush concepts, such as non-discrete brush boundaries, simultaneous display of multiple brushes, and creating composite brushes via logical operators. Preliinary experimental evaluation with test subjects supports the usefulness of N-dimensional brushes in data exploration tasks.

Source: IEEE Visualization '95 Conference