Enhancing diagnostic potential in language testing: A cognitive diagnostic modelling study through data-driven q-matrix refinement

Date

2025-9-22

Author

Yılmaz, Fahri

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

4194
views

0
downloads

Cite This

Individualised language instruction and learning require assessments that show which sub-skills an individual learner possesses and lacks. Cognitive diagnostic models (CDMs) can offer such feedback, but their validity depends upon an accurately specified Q-matrix associating items to latent attributes. Q-matrices constructed solely by subject-matter experts can be susceptible to bias, and inconsistent granularity potentially leading to inaccurate parameter estimation, examinee classification, and diagnostic inferences. This dissertation sought to develop an empirically informed Q-matrix construction procedure for L2 test development and employ CDMs for longitudinal development tracing of L2 sub-skills. A proposed data-driven, iterative Q-matrix refinement procedure under the G-DINA framework integrates CTT/IRT analyses for data preprocessing, exploratory factor structure discovery, and discrimination-index–driven refinement to inform expert judgements for Q-matrix construction. Its performance is evaluated against two methods in four simulations representing different noisy assessment conditions and scenarios. Substantively, a longitudinal case study design is used to compare the

Subject Keywords

Cognitive Diagnostic Modelling, Q-matrix Refinement, L2 Assessment, Longitudinal Learning Trajectories, G-DINA

URI

https://hdl.handle.net/11511/115678

Collections

Graduate School of Social Sciences, Thesis