@article{oai:nagoya.repo.nii.ac.jp:00001987,
 author = {水野, 欽司 and MIZUNO, Kinji},
 journal = {名古屋大學教育學部紀要. 教育心理学科},
 month = {Mar},
 note = {The purposes of this paper were to examine on the effectiveness of the following clustering procedure for contingency table, and to apply this procedure to the stratification of 174 school zones in Nagoya City based on the Census data. The procedure is summarized as follows : Concerning the frequency distribution of item Y (table-head item), the mutually similar categories of item X (table-side item) are combined to the same group under the condition of minimizing the loss of transmitted information T (x : y). Consequently, initial contingency table is reduced to a smaller one, which has several distinct groups combining similar categories of item X. This procedure can be easily extended to the case of multiway contingency table. Two algorithms, hierarchical and non-hierarchical (so-called hill-climbing), were examined against two kinds of Census data (1970), the educational distribution and the occupational one of Nagoya habitants, by school zone. Each school zone corresponds to each category of item X. The results of the computation with educational data showed five characteristic groups ordered one-dimensionally from 'high-educated' to 'low-educated'. Similarly, five groups were distinguished in the occupational case. These were 'residential zones in the surburbs', 'commercial zones', 'industrial zones' and so on. In both educational and occupational data, the achieved criteria T (x : y) by two algorithms were not so much different from each other. Larger difference between the two algorithms was recognized in the contents of school zones in each group at the occupational case, in which the loss of transmitted information was greater than the other. On the whole, it was found that both these clustering methods and the data of this kind might be useful for the stratification of school zones. However, whether such stratification is actually valid for various educational surveys or not is in need of further studies., 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。},
 pages = {1--15},
 title = {<原著>教育調査のための学区の層別},
 volume = {20},
 year = {1974}
}