Cluster analysis of age-grouped suicide data in Japan with spatio-temporal structure

摘要

Cluster analysis was applied to the spatio-temporal suicide data in Japan. The objective of the analysis was to detect the characteristics of clusters and to compare the results among age groups. The data is divided in terms of area, year, age group and sex. The cluster analysis was applied by the following procedure. First, the silhouette width was used to detect the appropriate number of clusters. Second, a hierarchical cluster analysis was applied to the data. Third, a nonhierarchical cluster analysis, k-means clustering, was applied to the data. Also, the meanings of the clusters, which were determined by the results, are explained by the background in Japan. Finally, the contrasts are discussed by comparing the initial terms of area, year, age group and sex.