A Projection Pursuit Method on the Multidimensional Squared Contingency Table

Ju Sun Ahn, Heike Hofmann, Dianne Cook

Research output: Contribution to journalArticleResearchpeer-review

5 Citations (Scopus)


In this study a projection pursuit method in used to explore ed (square) contingency table data. The method operates on projection matrices constructed from the contingency tables using affine geometry and creates projections (or marginals) using a Radon transform. The projection matrices and the projections can be used to find the "interesting" (nonuniform structure), and to cluster and to order the cases. This projection pursuit method is implemented with graph visualization of projection. It is similar to the discrete version of Andrews' curve. We demonstrate how this approach compares to association rules commonly used in data mining using a market basket data set and compare the PP results with the analysis of a data set from Wishart and Leach (1970).

Original languageEnglish
Pages (from-to)605-626
Number of pages22
JournalComputational Statistics
Issue number4
Publication statusPublished - 1 Jan 2003


  • Association rules
  • Data mining
  • Parallel class
  • Projection matrix
  • Projection pursuit
  • Radon transform
  • Squared contingency table

Cite this