Multiple scaling according to the principle of crystallization -: An alternative method to exploratory factor analysis

This paper focuses on one of the central methodological problems of scaling in the social sciences: the identification of multidimensional structures. In most practical research contexts exploratory factor analysis is used to solve this problem. However, the problem with this statistical procedure is that - due to certain restrictions - it does not always produce the desired outcome. Drawing evidence from one example, the author shows that under certain conditions complex dimensions are "fragmented" by orthogonal exploratory factor analysis. An alternative method which does not have distorting restrictions is a multiple scaling procedure for dichotomous items developed by Mokken (1971). Mokken scaling operates according to the principle of crystallization and reliably identifies multidimensional structures. Multiple scaling is a general multidimensional scaling method that, with the help of cluster analysis, can be implemented within classical test theory.