Answers are grounded in this paper and cite their sources.
Large-scale assessments (LSAs) primarily support system-level monitoring, but their instructional diagnostic potential remains underused. Cognitive Diagnostic Models (CDMs) offer a promising avenue, though their application in LSAs poses theoretical and practical challenges. This study explores whether cognitive models used in item generation can directly inform Q-matrix construction for CDM analyses, enhancing diagnostic value in early numeracy assessment. Using data from Luxembourg’s school monitoring program (N = 35,058), we analyzed four cognitive attributes (counting, addition < 10, decomposition, addition > 10) using developmental frameworks. We compared a Single-Attribute Hierarchical Model, assuming linear progression, with a Multiple-Attribute Hierarchical Model, allowing skill interactions. Both hierarchical models reproduced expected developmental progressions, with decomposition emerging as a key threshold skill and socio-economic status showing the largest subgroup differences. Subgroup analyses revealed a smaller-than-expected impact of migration background, while math anxiety peaked at intermediate skill levels. Embedding cognitive models in LSAs can bridge system-level monitoring and instructional support.
No takes yet. Share an insight, caveat, or question.
Sonnleitner et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: