RnavGraph: A Visualization Tool for Navigating through High-Dimensional Data Spaces
Adrian Waddell, Wayne Oldford
Statistics and Actuarial Science, University of Waterloo, ON, Canada

RnavGraph is an R package that facilitates controlled exploration of high dimensional data space via (user determined) low dimensional trajectories through that space. The trajectories are paths on a navigation graph (navGraph), a graph whose nodes are plots and whose edges represent transitions from one plot into another (see Hurley and Oldford, 2011, Comp. Stat.). In RnavGraph, there are two primary display regions – the navGraph in one and the “plot” (some visualization of the data) in the other. The navGraph display drives the data visualization. The user moves a “plot” circle, or “plot”, from one node to another along defined edges, causing one data visualization to be smoothly morphed into the other.

We give an overview of RnavGraph and demonstrate its functionality on some high dimensional data (e.g. images). The data visualization shown will be a 2d scatterplot of the high dimensional data objects (points, images, text, glyphs) specially implemented in Tcl/Tk to permit selection, zooming, and brushing. Each node of the navGraph corresponds to a scatterplot on two variates of the data; each edge represents either a 3d or 4d transition from one scatterplot to the other (depending on whether the node scatterplots share one variate or no variates). In the case of a 3d transition navGraph, moving the bullet from one node to the another simply rotates one scatterplot to another and a walk on this navGraph produces a display sequence of 2d scatterplots rotating into one another. As a roadmap for 3d movement through the higher dimensional space, it will be of interest to determine, save, and re-run interesting paths; the facility for doing so will also be shown.

RnavGraph also interfaces with Rggobi. We also show RnavGraph's design can provide the navigational infrastructure to build new visualizations.

Keywords: Data visualization; Graph navigation; R package; High dimensional data

Biography: Adrian Waddell is a doctoral student in Statistics under the supervision of Prof. Wayne Oldford at the University of Waterloo, Canada. Mr. Waddell holds an M.Math. (Statistics) from the University of Waterloo and a B.Sc. from the Zurich University of Applied Sciences.