clustering of rows and columns, e.g. in r by c contingency table, to detect patterns and aid understanding

Discussed in classics such as Jacques Bertin's 1967 Semiology of graphics and Helmut Spaths Cluster analysis algorithms (which also describes Walter D. Fisher's discretization algorithm implemented in XLStat Fisher procedure) such matrix reordering would allow understanding of r by c contingency tables, especially when r and c are greater than 2, e.g. which rows and columns cluster together. Otherwise researchers need to struggle with multinomial logistic regression etc or lots and lots of pairwise comparisons. Genetic algorithm version discussed by Niermann S Optimizing the ordering of tables with evolutionary optimization, American Statistician, 2005, 59, 41-46

  • Guest
  • Jun 13 2020
  • New ideas
  • Attach files