About me
I am a Professor of Psychological Assessment at the University of Kassel. My research focuses on psychological assessment, psychometrics, intelligence, educational measurement, and computational methods in assessment. Across these areas, my goal is to better understand how cognitive abilities can be measured reliably and efficiently, and how assessment can be improved through sound measures and better computational approaches.
Together with colleagues, I am currently working on the PINGUIN project, which develops adaptive tablet-based screening tools for elementary school children, and on the Meta-REP ML project, which examines the methodological rigor and robustness in applied machine learning research.
Resources
Tools & Code
Statistical tools, reproducible examples, code, and practical resources.
Tests & Questionnaires
Open tests, questionnaires, and measurement materials.
Selected Publications
Speck, K.-L., Jankowsky, K., Scharf, F., & Schroeders, U. (2026). Beyond the hype: A simulation study evaluating the predictive performance of machine learning models in psychology. Accepted for publication in Psychological Methods. https://doi.org/10.1037/met0000832
This simulation study shows that machine learning models in psychological research often do not live up to expectations and are constrained by the same basic data limitations as traditional regression models, especially small sample sizes, low predictor reliability, and small effect sizes.
Schroeders, U., & Achaa-Amankwaa, P. (2026). Developing NOVA: A next-generation open vocabulary assessment. European Journal of Psychological Assessment. Advance online publication. https://doi.org/10.1027/1015-5759/a000937
The article describes the development of NOVA—a modern open-access German vocabulary test developed as a transparent alternative to proprietary outdated measures. Moreover, item difficulty can be predicted to a considerable extent from word frequency and word length.
Schroeders, U., Wilhelm, O., & Olaru, G. (2016). Meta-heuristics in short scale construction: Ant Colony Optimization and Genetic Algorithm. PLOS ONE, 11. Article e0167110. https://doi.org/10.1371/journal.pone.0167110
The article contrasts conventional scale-shortening methods, which typically remove items stepwise based on reliability, with Ant Colony Optimization and Genetic Algorithms. These metaheuristics can optimize several psychometric criteria simultaneously, thereby helping to avoid the attenuation paradox, in which short forms are highly reliable but less valid.