SAS Institute Japan, JMP Japan Division Tester Yusuke Ono
This presentation will provide an overview of the geometric interpretation of principal component analysis and multiple correspondence analysis. From the perspective of French data analysis, which follows on from Benzecri, principal component analysis, correspondence analysis, and multiple correspondence analysis are statistical methods that project points scattered in a multidimensional geometric space into a lower-dimensional space. On the other hand, in Japanese how-to books for beginners, principal component analysis is often introduced as a method for constructing synthetic variables. Also, quantification type III is well-known in Japan as a method that applies the same calculations as multiple correspondence analysis, but quantification type III is often introduced as a method that maximizes the correlation between individual scores and category scores. These perspectives are somewhat different from the ideas in French data analysis.
The reports for principal component analysis and multiple correspondence analysis in JMP contain a mixture of French (geometric interpretation) and mainstream (composite variable and score interpretation) results. It seems necessary to have multiple perspectives when using these features in JMP.
This presentation will briefly introduce the French perspective on data analysis on principal component analysis and multiple correspondence analysis.
[References] Le Roux and Rouanet (2010) “Multiple Correspondence Analysis”, SAGE.
