Prof. Dr. Philipp Doebler, TU Dortmund University
Prof. Dr. Markus Gebhardt, TU Dortmund University
Jeffrey DeVries, TU Dortmund University
Children in at-risk circumstances with resilience have normal developmental outcomes in many areas, including education and cognitive development; meanwhile, those who lack resilience lag behind their peers. Several longitudinal studies have tracked the development of at-risk children in the resilience framework, but high quality studies including German children are lacking. The National Education Panel Study (NEPS) offers some unique features for both the study of resilience and educational outcomes with longitudinal data and methodological advances in test theory.
NEPS includes large-scale longitudinal data of several at-risk groups as well as not at-risk children in Germany. This data allows for identification and comparison of three at-risk groups of particular interest to current research: low socioeconomic status (SES) families, children with migrant background, and learners with special educational needs (SEN). Additionally, the large dataset of NEPS allows for a refinement of methodology for analyzing test items. Because of demographic and other differences in at-risk populations, individual test items may not provide the same information for both at-risk and not at-risk children (i.e., they lack measurement invariance). This would further complicate the problem of accurate comparing educational and cognitive development outcomes for at-risk children.
We propose to address these problems by identifying three at-risk groups including low SES, migrant background, and learners with SEN in the NEPS dataset and analyze their development longitudinally via competencies in math, science, and spelling as well as metacognition. We will use a novel differential item function (DIF) analysis to quantify the extent to which NEPS items are not invariant across both at-risk and not at-risk children. Finally, using plausible values we will produce quantitative longitudinal models of development, such as latent growth, profile, and change models for both groups to identify factors that predict resilience in at-risk learners.