Modeling and predicting non-linear changes in educational trajectories
- Funding agency: RWTH Start-Up Programm
- Project title: Modeling and predicting non-linear changes in educational trajectories: The multilevel latent growth components approach
- Applicant: Prof. Dr. Axel Mayer
- Staff members: Christoph Kiefer, M. Sc. & Caroline Keck, B. Sc.
- Duration: 12 months, 2017-2018
The investigation of developmental trajectories is a central goal in the social and educational sciences. However, investigating a research question like “Does a change of major after the first year of studies improve students’ satisfaction with their academic success sustainably?” already requires a statistical model that considers a) the potentially non-linear trajectories of satisfaction as well as b) the multiple levels of analysis (i.e. the individual level, the university level etc.) and c) a measurement model as satisfaction is a non-observable, latent variable. In this project, we use data from the National Educational Panel Study (NEPS) to develop and illustrate a new statistical approach that can adequately address all three challenges simultaneously. The new approach is termed multilevel latent growth components model (ML-LGCoM) and can be used to model the trajectory of an outcome-of-interest in its full complexity. A key feature of this approach is, that the functional form of the trajectories does not have to be pre-specified – in contrast to latent growth curve models. Thereby it allows researchers to adequately model the complexity of individual educational trajectories.
In our NEPS example, we model the (non-linear) development of students’ satisfaction with their academic success over four years and examine cluster- and individual-level trajectories. The new statistical model is broadly applicable and helps researcher to gain a more detailed view of human development over the life span in general.
Kiefer, C., Rosseel, Y., Wiese, B. S. & Mayer, A. (2018). Modeling and predicting non-linear changes in educational trajectories: The multilevel latent growth components approach. Psychological Test and Assessment Modeling, 60(2), 189-221.