Construct validity of the scales for rating the behavioral characteristics of superior students : a confirmatory factor analysis of internal structure

The purpose of the study was to investigate the construct validity of the Scales for Rating the Behavioral Characteristics of Superior Students (SRBCSS). A confirmatory factor analysis was used to determine if items from the scales of Learning, Motivational, and Creativity Characteristics were able to identify separate homogeneous sets of behavioral characteristics of gifted and talented students.The subjects were 152 third- and fourth-grade students who had been identified as being potential candidates for inclusion in a program for gifted and talented students. SRBCSS forms were completed by regular classroom teachers. The sample was split to allow for cross-validation of results.Two nested three-factor models were established a priori. Maximum likelihood estimates were calculated using LISREL IV (Jorskog & Sorbom, 1978). Models were evaluated using the overall chi-square, residual matrices, the ratio of chi-square to degrees of freedom, increment of fit indices, and differences chi-square.The proposed structure of three-uncorrelated constructs was unable to reconstruct the obtained correlations. The three factors of Learning, Motivation and Creativity were necessary but not sufficient to describe the relationships among the scale items. Allowing the constructs themselves to be interrelated improved the amount of covariation among the items which could be explained. However, adding a fourth Nonconformity Factor provided a more plausible representation of the data than either three-factor solution.However, labeling of the factors may be somewhat ambiguous due to the nature of the factor loadings of the individual items on them. Cross-validation showed a similar pattern of coefficients for the second sample with some "shrinkage" effect.The results of this study suggest that the validity of the constructs of the SRBCSS as establishedin is questionable. While separate dimensions emerged, they were neither easily interpretable nor distinctly separate. It was not clear if the interrelationships of the scale items were due to a sharing of some underlying dimension or to their placement on the scales. In addition, it was suggested that some higher order factor might be needed to account for the relationships found in the data.Thesis (D. Ed.