Exploring data through archetypes

Abstract

In this paper we propose a mixed analytical and graphical exploratory strategy based on data archetypes for the exploratory analysis of multivariate data. Our approach is of considerable help in exploring the periphery of the data scatter, exploiting an outward-inward perspective, to highlight small peripheral groups as well as anomalies, outliers and irregularities in the data cloud shape. The strategy is carried out in a comprehensive quantitative programming environment provided by the joint use of the software system R and of the visualization system GGobi. It provides a visualization system involving both static and dynamic graphics based on the so-called multiple views paradigm. The views are organized in a spreadplot and heavily exploit dynamics and interactive statistical graphics. © 2010 Springer-Verlag Berlin Heidelberg.