Premium
ALTERNATIVE FACTORIAL SOLUTIONS AND URBAN SOCIAL STRUCTURE: A DATA ANALYSIS EXPLORATION OF CALGARY IN 1971
Author(s) -
Vies W.K.D. Da
Publication year - 1978
Publication title -
canadian geographer / le géographe canadien
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.35
H-Index - 46
eISSN - 1541-0064
pISSN - 0008-3658
DOI - 10.1111/j.1541-0064.1978.tb01523.x
Subject(s) - rigour , objectivity (philosophy) , computer science , set (abstract data type) , field (mathematics) , dimension (graph theory) , consistency (knowledge bases) , data science , management science , operations research , epistemology , mathematics , artificial intelligence , engineering , philosophy , pure mathematics , programming language
I n recent years the development of the phenomenological (Stewart 1974) and prescriptive (Harvey 1973) approaches to the study of urban areas has increased the number of descriptive options open to the urban geographer. Undoubtedly these new methods have deepened our understanding of the complexity of urban areas, but it can be argued that these approaches are of little use to those with a positivist viewpoint who are concerned with providing consistent taxonomies or descriptions of urban character. Their basic problem is one of ensuring that the same set of descriptions or results would be obtained by any number of different investigators using the same data set. In studies of the dimensionality of urban social areas this requirement of consistency has led to factor analysis methods becoming the standard technical procedure in the field, primarily because these methods appear to provide the necessary objectivity and rigour needed by urban dimension and classification schemes. Moreover, they have the additional advantage of providing a choice between two distinct approaches: that of testing the adequacy of previous ideas or theories via common factor methods, and that of simply describing the maximum amount of variability of a data set by means of the component model. Unfortunately, the first option has rarely been used by urban ecologists, and despite the range of factor analysis procedures that are available (Harman 1976) practically all urban researchers have used one basic approach, namely the principal axes technique and component model (Clark, Davies, and Johnston 1974). It is well known that factors can be obtained from a given correlation matrix in a variety of ways so this restriction of effort to one procedure could mean that descriptions of the dimensionality of cities are technique‐dependent, in the sense that different factorial procedures could produce substantially different results from a single data set (Berry 1971; Hunter 1972). If these descriptions are technique‐dependent then any comparison between the factorial results of two studies of urban areas in which different factor procedures have been used can produce spurious conclusions.