Discussion by R. Gnanadesikan
The author of this well-written and interesting paper is clearly that rare combination of someone at the interface between a substantive science (astronomy in this case) and statistics who feels comfortable with the languages, problems, and methodologies of both domains. Indeed, on balance, the paper’s contents are tilted toward multivariate statistical data analysis methods (perhaps motivated by a desire to communicate with statisticians in the audience) and exhibit a knowledge of not only the more classical techniques but also more recent developments. The author’s knowledge is far from superficial on these matters and the paper demonstrates a critical ability to sort through the methods and raise relevant questions about their value and limitations.
KeywordsUnsupervised Classification Euclidean Representation Critical Ability Principal Curf Multivariate Data Analysis Method
Unable to display preview. Download preview PDF.
- deSoete, G. (1986). Optimal variable weighting for ultrametric and additive tree clustering. Quality and Quantity 20. 169–180.Google Scholar
- Gnanadesikan, R., Kettenring, J. R., and Tsao, S. L. (1990). Identification and selection of variables for cluster analysis. Invited talk presented at AERA meetings in Boston.Google Scholar
- Gnanadesikan, R. Harvey, J., and Kettenring, J. R. (1991). Metrics for sensitive cluster analysis. Work in progress.Google Scholar
- Lichtenstein, C. (1985). Discriminant Analysis when the Group Labels Are Uncertain. Ph.D. thesis, Cornell Univ., Ithaca, N.Y.Google Scholar