Statistical Methods in Transdisciplinary Educational Research

Abstract

Read online

A central task of educational research is to examine common issues of teaching and learning in all subjects taught at school. At the same time, the focus is on identifying and investigating unique subject-specific aspects on the one hand and transdisciplinary, generalizable effects on the other. This poses various methodological challenges for educational researchers, including in particular the aggregation and evaluation of already published study effects, hierarchical data structures, measurement errors, and comprehensive data sets with a large number of potentially relevant variables. In order to adequately deal with these challenges, this paper presents the core concepts of four methodological approaches that are suitable for the analysis of transdisciplinary research questions: meta-analysis, multilevel models, latent multilevel structural equation models, and machine learning methods. Each of these approaches is briefly illustrated with an example inspired by the interdisciplinary research project FALKE (subject-specific teacher competencies in explaining). The data and analysis code used are available online at https://osf.io/5sn9j. Finally, the described methods are compared, and some application hints are given.

Keywords

• transdisciplinarity
• meta-analysis
• multilevel model
• linear mixed model
• structural equation model
• machine learning