Reexamining ADHD-Related Self-Reporting Problems Using Polynomial Regression.

Individuals with attention deficit hyperactivity disorder (ADHD) underreport symptoms compared with informants and objective measures. This study applied enhanced statistical methodology (polynomial regression) to the study of ADHD self-reporting to clarify what contributes to symptom underreporting by adolescents with ADHD ( N = 107; ages = 11-15 years). Polynomial regression models were conducted to test competing hypotheses about the nature of self-reporting problems. Traditional difference score models were nested within polynomial regression models to examine how modeling strategy influences results. Sixty-six percent of the sample substantially underreported symptoms compared with parents and 23.6% denied all symptoms. Polynomial regression models provided no evidence that the size of the discrepancy between parent and adolescent symptom reports possessed meaningful linear associations with any of the hypothesized predictors. Nested models indicated that the difference score approach led to very poor model fit and increased risk for Type I errors when examining underreporting among youth with ADHD. This finding suggests that past evaluations using a difference score approach should be replicated using polynomial regression to ensure that significant effects do not represent statistical artifact.