In a recent JMR article, the authors described and illustrated a correspondence analysis scaling that permitted both within-set and between-set squared distance comparisons. This note clarifies the relationship between the proposed scaling and another scaling (popular among the “French school” of data analysts) in which the diagonal matrix of singular values is applied to both X and Y coordinates. Using a data set from Greenacre, the authors examine conceptual and empirical contrasts among three alternative scalings.
Get full access to this article
View all access options for this article.
References
1.
CarrollJ. Douglas, GreenPaul E., and CatherineM. Schaffer (1986), “Interpoint Distance Comparisons in Correspondence Analysis,” Journal of Marketing Research, 23 (August), 271–80.
2.
GifiA. (1981), Non-Linear Multivariate Analysis. Leiden, The Netherlands: Department of Data Theory, University of Leiden.
3.
GowerJ. C. and DigbyP. G. N. (1981), “Expressing Complex Relationships in Two Dimensions,” in Interpreting Multivariate Data, BarnettV., ed. Chicester, UK: John Wiley & Sons, Inc., 83–118.
4.
GreenacreMichael J. (1984), Theory and Application of Correspondence Analysis. London: Academic Press, Inc.
5.
HoffmanDonna L. and GeorgeR. Franke (1986), “Correspondence Analysis: Graphical Representation of Categorical Data in Marketing Research,” Journal of Marketing Research, 23 (August), 213–27.
6.
LebartLudovic, AlainMorineau, and KennethM. Warwick (1984), Multivariate Descriptive Statistical Analysis: Correspondence Analysis and Related Techniques for Large Matrices. New York: John Wiley & Sons, Inc.
7.
van derHeijden and Jan deLeeuw (1985), “Correspondence Analysis Used Complementary to Loglinear Analysis,” Psychometrika, 50 (December), 429–48.
