The Clustering of Categorical Data: A Comparison of a Model-based and a Distance-based Approach
- 27 January 2014
- journal article
- research article
- Published by Taylor & Francis in Communications in Statistics - Theory and Methods
- Vol. 43 (4) , 704-721
Abstract
For clustering multivariate categorical data, a latent class model-based approach (LCC) with local independence is compared with a distance-based approach, namely partitioning around medoids (PAM). A comprehensive simulation study was evaluated by both a model-based as well as a distance-based criterion. LCC was better according to the model-based criterion and PAM was sometimes better according to the distance-based criterion. However, LCC had an overall good and sometimes better distance-based performance as PAM, although this was not the case in a real data set on tribal art items.Keywords
This publication has 12 references indexed in Scilit:
- How to Find an Appropriate Clustering for Mixed-Type Variables with Application to Socio-Economic StratificationJournal of the Royal Statistical Society Series C: Applied Statistics, 2013
- Cluster Analysis2011
- Identifiability of parameters in latent structure models with many observed variablesThe Annals of Statistics, 2009
- Deciding on the Number of Classes in Latent Class Analysis and Growth Mixture Modeling: A Monte Carlo Simulation StudyStructural Equation Modeling: A Multidisciplinary Journal, 2007
- Clustering criteria for discrete data and latent class modelsJournal of Classification, 1991
- Finding Groups in Data1990
- Factor Analysis and AICPsychometrika, 1987
- Comparing partitionsJournal of Classification, 1985
- Exploratory latent structure analysis using both identifiable and unidentifiable modelsBiometrika, 1974
- Objective Criteria for the Evaluation of Clustering MethodsJournal of the American Statistical Association, 1971