An Empirical Assessment of Two Univariate Screening Measures in Cluster Analysis
As researchers soon discover, the inclusion of noisy (irrelevant) variables in cluster analyses can obscure or distort Atrue@ subgroup structures. This problem, identified and discussed by Milligan (1980), has prompted the search for methods that identify noisy variables and either down-weight or remove them. Several researchers have investigated this problem and have met with limited success (DeSarbo, Carroll, Clark, and Green 1984, De Soete 1986, 1988). Recently, Donoghue (1995) and Carmone, Kara, and Maxwell (forthcoming, 1999) have proposed screening methods to identify and eliminate noisy variables.
KeywordsCluster Structure Variable Weighting Heuristic Identification Multivariate Behavioral Research Subgroup Structure
- Carmone, Jr., Frank J., Kara, Ali, and Maxwell, Sarah(1999), AHINoV, A New Model to Improve Market Segment Definition by Identifying Noisy Variables, @ forthcoming.Google Scholar
- Carroll, J, Douglas (1973), AHoward-Harris Clustering,@ Appendix B in P.E. Green, and Y. Wind, Multiattribute Decisions in Marketing, Hindsale, IL: Dryden Press, 368-71.Google Scholar
- Cliff Norman (1992), PAIRDEL1.BAS. Program for Computing Matched-data D-statistics [computer program]. Los Angeles: Psychology Department, University of Southern California.Google Scholar
- SAS (1985), SAS User=s Guide: Statistics, version 5 edition. Cary, NC: AuthorGoogle Scholar