Exploratory Data Analysis of Multivariate Data
using Interactive Dynamic Graphics
Dianne Cook
Department of Statistics, Iowa State University
Statisticians have long struggled with how to display multiple
variables on a 2D sheet of paper, or 2D computer screen. In a 1985
paper, Dan Asimov discussed the grand tour as a way to "see
multivariate data from all sides". He proposed to create animations
that show views of p-dimensional data as seen through the windshield
of a helicopter flying in p-space. Asimov's original grand tour was
essentially on auto-pilot, with no facilities for human intervention.
Since then many efforts have been made to put a human pilot in charge
of the grand tour flight path. Some of these efforts borrow from
multivariate analysis, projection pursuit, Andrews curves, parallel
coordinates, and 3-D trackball controls. In this talk we will discuss
the use of the grand tour, and the most recently invented controls,
for pattern recognition and structure detection in multivariate data.