Search

A previous paper (Hinings et al. 1967) described an approach to the study of the structure of bureaucratic organizations based on the creation of scaleable dimensions for multivariate analysis. The present paper considers in detail the process of the scaling of an organizational variable and the use of two multidimensional analyses, namely, factor analysis and multiple regression analysis. These are illustrated by analysis of the operationally defined concept of functional specialization. The central points made are: (i) that any data analysis is a function of a theory of organizational behaviour, an error theory, and a computing algorithm; and (ii) that the choice of a method of analysis necessarily implies a choice of a theory of organizational behaviour. In the case of scaling, Guttiman's scalogram method and item analysis, borrowed from psychological test theory, are compared. They are found to differ more in computing technique and in their representation of error than in their measurement theory. It is pointed out that the multivariate analyses employed are both compensatory theories of organizational behaviour; the implications of this are discussed. [ABSTRACT FROM AUTHOR]

Most efforts to analyze and forecast interregional population growth and distribution are hampered by the general paucity of reliable data on interregional migration. This has led demographers to rely on crudely constructed measures of net migration. Recent efforts to express the population growth process in matrix form, however, suggest a means whereby the schedule of growth that operates in an interregional system may be estimated solely on the basis of historical data on interregional population distributions. Such a method is described and illustrated in this paper. The method estimates a growth regime which, when applied to the initial interregional population distribution in the time series, most closely reproduces the remaining population distributions in the historical sequence.