Editorial

Abstract

Introduction

The Projection Pursuit

RADVIZ

Parallel Coordinates

User Interaction

Finding Outliers

Conclusions

References

Appendix A - Datasets

Appendix B

Case Studies Index

An Investigation of Methods for Visualising Highly Multivariate Datasets

Editorial Introduction

The date used by social scientists are frequently multivariate. In part, this is a consequence of a need to characterise objects of interest, such as people, houses and so on, as fully as possible, but it is also often a result of a desire to capture concepts such as social class or intelligence and overcrowding that do not permit easy measurement along one axis of variation. In consequence, quantitative social science has a long history of using statistical and mathematical transforms of data matrices such as factor and principal component analysis to reduce the dimensionality of these data and perhaps suggest appropriate constructs that might also be used to describe individuals.

These techniques are not intrinsically visual, although the reprojection of individual cases onto axes that define these constructs (for examples as component scores) may well create data that can be visualized by any of the standard techniques. There remains a need to develop appropriate alternative visualizations for multidimensional data that are efficient in allowing the detection of patterns in the multivariate data space. In this Case Study, Chris Brunsdon, Stweart Fotheringham and Martin Charlton develop and illustrate three alternative projections that can be applied to multivariate data.

It is interresting to note that although the static displays produces are in themselves useful, they gain maximum utility when visualized in an interactive environment.