2024-04-21T08:47:43Z
https://nagoya.repo.nii.ac.jp/oai
oai:nagoya.repo.nii.ac.jp:00001454
2023-11-09T03:02:44Z
323:350:351:352
Applications of Nonmetric Principal Components Analysis forCategorical Data
カテゴリカル・データの非計量的主成分分析の応用
村上, 隆
3408
Murakami, Takashi
3409
categorical data
principal component analysis
nonlinear multivariate analysis
quantification of qualitative data
multiple correspondence analysis
1997-12-26
An extended method of nonmetric (nonlinear) principal components analysis (PCA) for (unordered) categorical data, which gives multiple quantifications to categories of an item maximizing the size of explained variances by the factor analytical model with the specified number of components, is applied to three data sets; ratings of 34 drinks on 7 categorical items, choices of the most important agent from the list including 9 candidates by 274 experienced persons in terms of promoting developments of local districts on 17 domains, and self ratings of 1630 university students on 10 self-esteem items with 4 ordered categories. The main findings are as follows : (1) multidimensional quantifications of an item by the new method described some interesting features of drinks in the first data set which could not be found by the unidimensional quantification. (2) Nonmetric PCA gave a compact and interpretable representation of the five dimensional structure of nominal variables in the second data set. (3) Relationships between results of Multiple Correspondence Analysis of ordered categorical variables, which is an unconstrained version of nonmetric PCA, and those of classical PCA treating the same variables as interval scales were clarified by nonmetric PCA of the third data set. (4) Rotational freedom of the model and availability of the matrix of loadings in the method facilitated interpretations greatly through the three cases of applications.
国立情報学研究所で電子化したコンテンツを使用している。
departmental bulletin paper
名古屋大学教育学部
1997-12-26
名古屋大學教育學部紀要. 心理学
44
87
105
http://hdl.handle.net/2237/2866
03874796
jpn