Chapter

Innovations in Classification, Data Science, and Information Systems

Part of the series Studies in Classification, Data Analysis, and Knowledge Organization pp 81-90

Measuring Distances Between Variables by Mutual Information

  • Ralf SteuerAffiliated withNonlinear Dynamics Group, University of Potsdam
  • , Carsten O. DaubAffiliated withMax-Planck Institute for Molecular Plant Physiology
  • , Joachim SelbigAffiliated withMax-Planck Institute for Molecular Plant Physiology
  • , Jürgen KurthsAffiliated withNonlinear Dynamics Group, University of Potsdam

* Final gross prices may vary according to local VAT.

Get Access

Abstract

Information theoretic concepts, such as the mutual information, provide a general framework to detect and evaluate dependencies between variables. In this work, we describe and review several aspects of the mutual information as a measure of ‘distance’ between variables. Giving a brief overview over the mathematical background, including its recent generalization in the sense of Tsallis, our emphasis will be the numerical estimation of these quantities from finite datasets. The described concepts will be exemplified using large-scale gene expression data and compared to the results obtained from other measures, such as the Pearson Correlation.