The Generalizability of Two Rating Scales of Executive Function Across Parent and Self-Reports in a School-based Adolescent Sample
Data is provided by the student.
This study examined the reliability, validity, and generalizability of test scores obtained from parents and adolescents reports across two rating scales of executive function. Forty-two adolescents ages 12 to 17 enrolled in a rural public school and one of their parents participated in the study. For each adolescent-parent dyad, 2 total scores of executive function and 6 subscale scores hypothesized to measure the same theoretical constructs were calculated and scores were submitted to several Generalizability theory analyses to evaluate the instrument effect, rater effect, dimension effect, and all interactions on total scores and subscale scores. The resulting dependability coefficients were markedly low (i.e., .16 to .69) and much lower than expected given prior research and the corresponding Pearson correlations evaluating these facets in isolation. The subject-by-rater interaction contributed the largest proportion of variance in test scores (i.e., 30% to 33%). Surprisingly, the dimension effect and its interactions contributed little variation in test scores, suggesting that the construct of executive function is unidimensional. Results indicated that most of the subscale score variance was due to the inclusion of the rater facet, and particularly to the adolescent raters. The results from this study emphasize the importance of using extreme caution when generalizing scores of executive function across different instruments, raters, and specific executive functions.