In this paper, first, similarity between categories is formulated using AIC. Then the relation between dual scaling solution and category pooling process conducted by AIC is discussed . Second, a method analyzing cross tabulated data is proposed, which is useful for handling multicategorized variables. In this method, categories are pooled into some groups so as to minimize AIC of cross table composed by the pooled category groups. Then, some tables that not significantly different from the MAICE table are selected, and several inductive results are inferred and confirmed by comparing dual scaling solutions on the selected tables and by analyzing the category pooling process. This confirmation procedure also provides deeper understanding of dual scaling solutio Applying the method to two cross tables of vehicle possession in Tukuba Science City, it is shown that proposed method is practically efficient for drawing conclusions, even in the cases several categories have rather small samples and the table apparently has no significance.
抄録全体を表示