Psychotic and neurotic depression: I. Some points of method.

Testing hypotheses on classification could be made more precise by using only clinical symptoms and signs as its basis. Analyses based on a mixture of clinical, aetiological and background variables give ambiguous results since they incorporate the assumption that aetiological and symptomatological patterns are related in a one-to-one manner. A set of simplified models of possible relationships between aetiology and symptomatology is used to discuss these problems of interpretation. The clinical variables to be used for discriminant analyses may be chosen by testing for the existence of lower-order and higher-order differences between groups.