Use of Correspondence Analysis in Clustering a Mixed-Scale Data Set with Missing Data

Abstract

Correspondence analysis is a method of dimension reduction for categorical data, providing many tools that can handle complex data sets. Observations on different measurement scales can be coded to be analysed together and missing data can also be handled in the categorical framework. In this study, the method s ability to cope with these problematic issues is illustrated, showing how a valid continuous sample space for a cluster analysis can be constructed from the complex data set from the IFCS 2017 Cluster Challenge.