This page serves as a repository of software and source files developed by staff and students associated with the M&T group
Hierarchical cluster analysis as implemented in most of the well known statistical computer programs neglects the phenomenon of input order instability. That is, cluster solutions may differ when the rows and colums of the proximity matrix are permuted. This phenomenon is not widely known and is caused by ties that are present in the initial (dis)similarity matrix or arise during the process of clustering. To tackle this phenomenon, PermuCLUSTER has been developed.
PermuCLUSTER repeats the analysis a large number of times by permuting the rows and columns of the proximity matrix. In order to compare the solutions and find the optimal solution, a goodness-of-fit measure is used. The number of times the matrix should be permuted is variable and is user defined. PermuCLUSTER is an SPSS add-in and offers all but the same functionality as CLUSTER in SPSS.
- PermuCLUSTER18.104.22.168-Setup.exe zipped (updated oct 17, 2006)
- PermuCLUSTER 1.0 User's Guide (pdf)
- Rules of SPSS Hierarchical Cluster Analysis for processing ties (pdf)
- Completely elaborated example of data order instability in a Between-group Average hierarchical cluster analysis (pdf)
- For using PermuCLUSTER in SPSS 14.0 please read Advice for SPSS 14 Users (pdf)
MATLAB m-files to analyse three-way data and a short description.txt file are provided. Each slice of the three way box is decomposed by a singular value decomposition. The singular values or vectors may be constrained to be equal over the slices.
The procedure is described in De Rooij, M. and Anderson, CJ. Visualizing, Summarizing, and Comparing odds ratio structures. Manuscript under revision.
- MATLAB m-files + instruction (zipped)