Multimethod Assessment of Childhood Depression: Evidence for Convergent and Discriminant Validity Across Developmental Levels.

Abstract

The convergent and discriminant validity of three assessment methods were investigated in relation to the construct of childhood depression for a sample of elementary, intermediate, and secondary students. The three assessment methods of self-report, teacher, and peer rating scales were used to assess the response classes of depression, social withdrawal, social skills, and aggression. The self-report, teacher, and peer rating scales were subjected to a series of analyses to determine their psychometric properties. Test-retest and internal consistency reliabilities were generally acceptable, although the stability and item homogeneity of individual factors varied from the low-to-high range within each of the rating scales. Four factors (i.e., Depression, Social Skills, Aggression, and Social Withdrawal) were extracted from the self-report, teacher, and peer rating scales following a confirmatory factor analysis. Evidence for the criterion-related validity of the three rating scales was obtained using a self-report rating scale of depression, and two teacher rating scales of social skills, externalizing behavior problems, and internalizing behavior problems. The main purpose of the present study was to examine the convergent and discriminant validity of childhood depression across three sources and three grade levels through four multitrait-multimethod (MTMM) matrices. Evidence for convergent validity was found for the full MTMM matrix of grades 3 through 12 and for the three grade levels (i.e., 3-6, 7-9, and 10-12) using Campbell and Fiske's criteria. Of the 12 validity coefficients in each matrix, 10 attained statistical significance in the full matrix, 7 in the elementary grade matrix, 6 in the intermediate grade matrix, and 8 in the secondary grade matrix. Although the Aggression factor and the self/peer method showed the highest convergent validity, some evidence was obtained for the Depression factor, the Social Skills factor, and the other methods. Minimal evidence was gathered for discriminant validity using Campbell and Fiske's criteria. Results of an analysis of variance model for MTMM matrices showed significant convergent and discriminant validity for each of the four matrices, but also significant method variance. Findings were discussed in relation to the methodological problems in the assessment of childhood depression and to suggested areas of future research.