Projection Pursuit


The aim of exploratory projection pursuit (EPP) is to find linear combinations of variables so that data exhibit clusters with respect to those variables. EPP is similar in in some ways to principal components analysis.

Two-dimensional PP

An excellent package that (among other things) can perform two-dimensional projection pursuit is called XGobi and available from the StatLib archive. A version of Xgobi for the freeware R statistical software package is at CRAN

The software associated with Friedman, J.H. (1987) Exploratory projection pursuit. Journal of the American Statistical Association, 82, 249--266 is also available from the StatLib archive.

Chris Jones has kindly agreed to allow us to distribute his two-dimensional pp subroutines written in FORTRAN. He has written a report describing the software. Further details can be found in Jones, M.C. and Sibson, R. (1987). What is projection pursuit? (with discussion). Journal of the Royal Statistical Society, 150, 1--36.

Three-dimensional PP

There is a technical report about an implementation that performs projection pursuit into three-dimensions. A more recent published version may be found in Journal of the Royal Statistical Society, Series C, 44, 411-430.

This work was about projecting multi-spectral images into colour images like the three-colour composite of Chew Valley Lake in Somerset.

There is some free software for S that can perform three-dimensional EPP.

PhD Thesis

I did my PhD thesis [PDF] [ps.gz] on projection pursuit.

Functional projection pursuit

Functional projection pursuit is the marriage of functional data analysis and exploratory projection pursuit. This paper will appear in the Proceedings of the Interface, Houston, 1997.

Postgraduate opportunities

See the postgraduate opportunities page if you wish to pursue research in projection pursuit.

Publications

See my publications list.
G.P.Nason@bristol.ac.uk