CORRECT CLASSIFICATION RATES IN MULTIPLE CORRESPONDENCE ANALYSIS(Theory and Applications)

• Adachi Kohei

  Department of Psychology, Ritsumeikan University

Abstract

Multiple correspondence analysis (MCA) is formulated by various approaches, and homogeneity analysis (HA) is a major one among them. However, the HA approach has not yet provided a suitable index of GOF (goodness-of-fit) of multidimensional solutions. In this paper, the use of a correct classification rate (CCR) as the index is considered. We argue that CCR is congruous to the homogeneity assumption underlying HA, to justify the use of CCR in HA. Following this, we perform a simulation study to evaluate CCR by comparing it with eigenvalue-based GOF indices which have been derived from another approach to MCA. In the simulation CCR showed better performance than the eigenvalue-based indices: the former was found useful for evaluating the quality of MCA solutions and choosing solutions of proper dimensionalities. CCR also gave reasonable results in real data examples.

Journal

• Journal of the Japanese Society of Computational Statistics

  Journal of the Japanese Society of Computational Statistics 17 (1), 1-20, 2004-12

  Japanese Society of Computational Statistics

Keywords

• Goodness-of-fit indices
• Homogeneity analysis
• Multivariate categorical data
• Scree plots

Details

• CRID: 1573387449718260864
• NII Article ID: 10014459123
• NII Book ID: AA10823693
• ISSN: 09152350
• Text Lang: en
• Data Source:
  • CiNii Articles