大正・昭和戦前期における自動車の普及過程

4変量傾向面分析の適用による再検討

抄録

The present report announces the sequel of the report which appeared in this publication in 1988. The diffusion process of automobiles which had been used on the Japanese road traffic as modern transport facility, during the Taisho and the pre-war Showa eras (1912-1937) was discussed through descriptive analysis in the first report. This time we shall try to clarify the same problem as the first report using a quantitative method of four-variable trend surface analysis.
The study was made according to the following arrangements: First, the number of automobiles per ten thousand population was calculated for each of forty-seven prefectures for the year's 1915, 1920, 1925, 1930 and 1935. The distribution of these data was to be considered in two kinds of space. Namely, one is a physical space (geographical space), and the other is a two-dimensional Euclidean common space.
In order to recover the two-dimensional common space the individual difference scaling (INDSCAL), one of the multidimensional scaling was applied to the set of two dissimilarity matrices, elements of which are dissimilarity indices of both occupation structure between prefectures and birthplace structure between prefectures. As much as 88.9% of the variance in the matrices was explained by the two-dimensional common space (Fig. 2). The first co-ordinate axis was interpreted as occupation structure: prefectures with high commercial and manufacturing occupation rates lie on the left and those with high agricultural occupation rate on the right. The second co-ordinate axis could be interpreted as regionalization: eastern prefectures lie on the upper and western ones on the lower.
Linear through cubic four-variable trend analyses were applied to the number of automobiles per ten thousand population to examine the diffusion process in each space. The time-slices of quadratic hypersurf aces were drawn both in the physical space (Fig. 1) and in the common space (Fig. 3).