An application of Mokken's model for stochastic, cumulative scaling in psychiatric research

Rating scales used in psychiatric research are often based on factor analysis. In practice, the resulting factors or dimensions will generally be empirically derived rather than based on theoretical considerations. In addition, it often proves difficult to determine their validity. It is argued that the use of an hierarchical scale may be a partial answer to these problems. This is illustrated by the analysis of data from two interview schedules, using Mokken's model for stochastic, cumulative scaling. Data were derived from a follow-up study of patients with a first life-time episode of non-affective, functional psychosis. Two scales for the assessment of psychopathology and social dysfunctioning were constructed. Scale statistics proved to be satisfactory, while the order of items in the scales corresponded to a priori theoretical expectations.