[an error occurred while processing this directive]

Workshop:

the workshop is kindly supported by SPSS the workshop is kindly supported by StatSoft

" Statistical data mining between research and practice "

27./28. Februar 2004 in Hamburg

Hans-Hermann Bock
(Institut für Statistik, RWTH Aachen)


[ Main Page]   [ Program]   [ Speakers]   [ Participants]   [ Travel Information ]

Statistical Analysis and Classification

Whereas classical data show real-valued or categorical entries in the cells of a data matrix, these entries are interval-valued, multi-categorical, or even frequency distributions in the case of symbolic data. Moreover there might be structural or hierarchical dependencies between variables. Symbolic data analysis extends classical methods from multivariate statistics to the case of such generalized data tables. Whereas classical statistics is mainly based on probability models, the modified approach here is often based on geometrical or heuristical considerations.
This paper presents a survey on the statistical tools for symbolic data which were implemented in the software system SODAS and illustrates the methods by practical examples using SODAS.
More specifically, the paper comprises the following sections:
  • Symbolic data and 'symbolic objects'
  • Visualization tools (zoom star, descriptive statistics)
  • Principal component analysis for interval data
  • Dissimilarity measures for symbolic data (including, e.g., missing values)
  • k-means-type clustering methods
  • Hierarchical and pyramidal cluster analysis with 'complete objects'
  • Kohonen maps for symbolic data
  • Symbolic factorial analysis
  • Conclusions.

Reference:
Bock, H.-H., Diday, E. (2000): Analysis of symbolic data. Exploratory methods for extracting statistical information from complex data. Studies in Classification, Data Analysis, and Knowledge Organization. Springer, Heidelberg

 Seitenanfang  Impressum 20. Feb. 2004, von Stefan Heitmann