@article{oai:sapmed.repo.nii.ac.jp:00014514,
 author = {森, 満},
 issue = {4},
 journal = {札幌医学雑誌 = The Sapporo medical journal, The Sapporo medical journal},
 month = {Dec},
 note = {Japan yearly publishes several health and related indicators by 47 prefectures, which could be presented by a fewer number of groups using both hierarchical cluster analysis （HCA） and principal component analysis （PCA）. As no study involving both HCA and PCA, applied to the data set of heath indicators, was found in Japan, our study purposes were: （i） to determine fewer groups of indicators from the 40 health and related indicators by applying both methods, and then （ii） to compare their groups with each other. First HCA was applied to the data to have the dendrogram of 40 indicators, after that the dendrogram was analyzed by dendrogram sharpening technique to identify the smaller groups （clusters） of either 1 or 2 indicators for the exclusion purpose from further analysis. Remaining 30 indicators （after dropping 10 indicators by dendrogram sharpening） were regrouped by HCA and compared with the groups of PCA. Reanalyzing them, HCA identified five groups （clusters） which were labeled as “C1: health care facility and cause-specific mortality”,“C2: morbidity”,“C3: welfare opportunity”,“C4: overall mortality”, and “C5: social status”. Similarly PCA showed 5 groups （PCs） which explained 86% of the total variation. These were labeled as “P1: health care facility”, “P2: socio-economic standard and cause-specific mortality”, “P3: welfare opportunity”, “P4: morbidity”, and “P5: overall mortality”. Comparative results revealed C2=P4, C3=P3, and C4=P5, whereas remaining groups overlapped highly by indicators. This study revealed that after dendrogram sharpening, both HCA and PCA provided almost similar groups of indicators and hence indicated their applicability to the same set of data. Dendrogram sharpening also made the interpretation more understandable by dropping the smaller groups of indicators.},
 pages = {39--50},
 title = {日本における健康とその関連指標のグループ化 : 階層的クラスター分析と主成分分析の類似性},
 volume = {73},
 year = {2004}
}