Cross-Informant Agreement for Ratings for Social Skill and Problem Behavior Ratings: An Investigation of the Social Skills Improvement System-Rating Scales

• Published in: Psychological assessment 2010-03, Vol.22 (1), p.157-166
• Main Authors: Gresham, Frank M, Elliott, Stephen N, Cook, Clayton R, Vance, Michael J, Kettler, Ryan
• Format: Article
• Language: English
• Subjects: Adolescent; Adolescents; Behavior; Behavior Problems; Biological and medical sciences; Caregiver Attitudes; Caregivers; Child; Children; Convergent Validity; Correlation; Discriminant Validity; Female; Human; Humans; Informants; Interpersonal Competence; Male; Medical sciences; Multitrait Multimethod Techniques; Observer Variation; Parent Attitudes; Personality Assessment - statistics & numerical data; Psychological tests; Psychology. Psychoanalysis. Psychiatry; Psychometrics - statistics & numerical data; Psychometrics. Diagnostic aid systems; Psychopathology. Psychiatry; Quantitative psychology; Rating; Rating Scales; Reproducibility of Results; Social Behavior Disorders - diagnosis; Social Skills; Socialization; Teacher Attitudes; Techniques and methods; Test Validity; Validity
• ISSN: 1040-3590; 1939-134X
• DOI: 10.1037/a0018124

One of the most consistent findings in rating scale research with children and adolescents is the modest agreement among different informants' ratings. The present study systematically explored patterns of agreement among teachers, parents/caregivers, and students in domains of social skills and problem behaviors using the Social Skills Improvement System-Rating Scales (SSIS-RS; F. M. Gresham & S. N. Elliott, 2008 ). Two subsamples from the normative sample of the SSIS-RS were used. The first sample of participants consisted of 168 students who had all 3 informants (parent, teacher, and self) complete the SSIS-RS scales, which was necessary to assess agreement across different raters. The second sample consisted of 164 students who had raters in a similar or same role (father-mother, teacher-teacher). The results replicated an extensive literature showing that cross-informant agreements for social skills and problem behaviors are weak to moderate. The current study invoked multitrait-multimethod logic to interpret the correlations among raters derived from different informants and showed that the convergent validity coefficients were consistently stronger than the discriminant validity correlations. Implications for assessment practices and future research are discussed.