A consideration to Urban Analysis

Abstract

A great number of analytical schemes have been developed in the history of the studies of cities. These are divided into two types : qualitative analysis and quantitative analysis. In our own study we made an attempt to understand the characteristics of cities by means of the principal component analysis within the quantitative analysis. By this we think we were able to understand the relationships between many variables that characterize and differentiate the cities, thus minimizing the important stress of any preconceived order that the researcher might have, and maximizing the importance of the underlying order in the data. Thus we think we were able to find out the parsimonious set of variables that account for the complicated urban phenomena. We planned to cover all the cities in the Kinki area except those of Mie prefecture. Three separate analyses were made ; one with 63 cities and 24 variables in 1955, the second with 74 cities and 28 variables in 1960, the third with 75 cities and 28 variables in 1965. Although an attempt was made to include as wide a range of variables as possible relating to demographic, social and economic characteristics, the coverage was not complete. The three separate analyses yielded almost the same set of factors. Six common factors were extracted at each period, but over 70 per cent of the variance was explained by the first three factors only. Then we used these three factors as a basis for our study. Further we proceeded to the analysis by using factor scores, namely, we dealt with the tolerance limit of the factor scores and the characteristics of the cities, the relationship between factor scores and city groupings, the ideal city and actual city, and Markov chain model. Indeed, there may be several shortcomings and imperfections in this study, but these applications turned out to be useful for analyzing the characteristics of the cities. Furthermore, we will first of all classify the area to be studied into several classes by means of the latent profile analysis, and then carry out the principal component analysis in each class. Also, we intend to compare one class with another.