Diagnostic Classification Models: Recent Developments, Practical Issues, and Prospects

Abstract

More than three decades after their introduction, diagnostic classification models (DCM) do not seem to have been implemented in educational systems for the purposes they were devised. Most DCM research is either methodological for model development and refinement or retrofitting to existing nondiagnostic tests and, in the latter case, basically for model demonstration or constructs identification. DCMs have rarely been used to develop diagnostic assessment right from the start with the purpose of identifying individuals’ strengths and weaknesses (referred to as true applications in this study). In this article, we give an introduction to DCMs and their latest developments along with guidelines on how to proceed to employ DCMs to develop a diagnostic test or retrofit to a nondiagnostic assessment. Finally, we enumerate the reasons why we believe DCMs have not become fully operational in educational systems and suggest some advice to make their advent smooth and quick.

Keywords:

• attribute
• DCM application
• diagnostic classification models
• model fit
• Q-matrix

Funding

The second author has been supported by Alexander von Humboldt Foundation via Group Linkage Program.

Notes

1 Slips are aberrant responses. Simply put, slips are careless errors. Slipping occurs if a respondent who has mastered all the attributes required by a given item, slips and answers the item incorrectly.

2 Guesses are another type of aberrant responses. They are lucky guesses. Guessing occurs if a person provides a correct answer to the item although he has not acquired all the attributes measured by the item.

3 Originally, the NC-RUM has been parameterized as a noncompensatory model. However, Ma, Iaconangelo, and de la Torre (2016) showed that it is an additive DCM with log link function.