Abstract: | This paper attempts to further the investigation of the close relationship between the seemingly diverse statistical techniques of Guttman scaling and principal components (factor) analysis by showing that artificial data sets composed of various kinds and combinations of cumulative scales can be analyzed in a meaningful and related way by principal components analysis. The comparison is then moved to a more abstract level through the postulation of a definition of “similarity” or association between roll-calls which is used as a standard for evaluating the two techniques. The outcome is that principal components analysis is interpreted as a refinement over Guttman scaling in the assessment of roll-call similarity, although where this extra precision is not required, Guttman scaling remains an acceptable methodology. By raising the question of the relation of technique to data to this level, it is hoped that the bases for more theoretically oriented choices of technique have been established. |