Composite variable analysis: A simple and transparent alternative to factor analysis

Composite variable analysis is a simple clustering procedure, developed specifically for use with data derived from responses to single items. It provides useful information about the latent variable structure of psychological measures, which is very similar to that provided by factor or principal components analyses. The procedure is described in some detail, and an illustration of its use is given, using a data set derived from the responses of 258 subjects to a measure with which conventional factor analysis has considerable difficulty. Results are compared with those obtained from varimax rotated principal components analyses, and it is concluded that composite variable analysis provides a better basis for decisions related to the adequacy of items than traditional factor analytic procedures.