Confirmatory factor analysis and factorial invariance analysis of the adolescent self-report Strengths and Difficulties Questionnaire: How important are method effects and minor factors?

Design. Data were obtained from the Youth Health Monitor Rotterdam, a community-based health surveillance system. Methods. The sample consisted of 11,881 pupils of 11-16 years old. Next to the original five-factor model, a factor model with the number of factors based on parallel analysis and scree test was investigated. Confirmatory factor analysis for ordered-categorical measures was applied to examine the goodness-of-fit and factorial invariance of the factor models. Results. After allowing reverse-worded items to cross-load on the prosocial behaviour factor and adding error correlations, a good fit to the data was found for the original five-factor model (emotional symptoms, conduct problems, hyperactivity-inattention, peer problems, prosocial behaviour) and a model with four factors (emotional symptoms and peer problems, conduct problems, hyperactivity-inattention, prosocial behaviour). Factorial invariance across gender, age, level of education, and ethnicity was found for the final five- and four-factor model, except for the prosocial factor of the four-factor model that showed partial invariance across gender. Conclusions. While support was found for both models, the final five-factor model is theoretically more plausible and gained additional support as the original scales emotional problems and peer problems showed different relations with gender, educational level, and ethnicity.