Cross-validation and response sets in repeated use of mood questionnaires