×

Correspondence analysis versus principal component analysis for highly skewed distributed variables. (English) Zbl 0616.62081

In the last fifteen years a method of analysis has become known in France under the name ”Analyse des correspondences”. In geography this method is very well suited to representing spatial relationships between variables and their categories, usually called regions in this discipline.
In this paper it is demonstrated that correspondence analysis has large advantages over principal component analysis, especially for highly skewed distributed variables. Unlike principal component analysis, correspondence analysis is not sensitive to such deviations from normality.

MSC:

62H25 Factor analysis and principal components; correspondence analysis
62H99 Multivariate analysis
PDFBibTeX XMLCite