workshop

Cognitive Diagnosis Modeling: A General Framework Approach and Its Implementation in R

Updated: 9:16am, 16 Jun, 2026
Date:
27 July 2026 (Mon)
Time:
9:00am - 6:00pm
Venue:
Room 754, Meng Wah Complex, The University of Hong Kong
Speaker(s):
Target Audience:
Open to RPg and EdD Students, Teaching and Research Staff, and Assessment Professionals
Registration Fee:
Registration is free, but slots are limited
Recording:
Related Files:
Photo Highlights:

About the Workshop

The primary aim of the workshop is to provide participants with the necessary practical experience to use cognitive diagnosis models (CDMs) in applied settings. Moreover, it aims to highlight the theoretical underpinnings needed to ground the proper use of CDMs in practice.

In this workshop, participants will be introduced to a proportional reasoning (PR) assessment that was developed from scratch using a CDM paradigm. Participants will get a number of opportunities to work with PR assessment-based data. Moreover, they will learn how to use GDINA, an R package developed by the instructors for a series of CDM analyses (e.g., model calibration, evaluation of model appropriateness at item and test levels, Q-matrix validation, various classification accuracy indices). To ensure participants understand the proper use of CDMs, the theoretical bases for these analyses will be discussed.

The intended audience of the workshop includes anyone interested in CDMs who has some familiarity with item response theory (IRT) and R programming language. No previous knowledge of CDM is required. By the end of the session, participants are expected to have a basic understanding of the theoretical underpinnings of CDM, as well as the capability to conduct various CDM analyses using the GDINA package.

Workshop Objectives

  1. To provide an overview on the theoretical foundations of CDM
  2. To introduce the development of a cognitively diagnostic PR assessment PR
  3. To illustrate how an R package developed by the instructors can be used for various CDM analyses

Importance of the topic

Unlike unidimensional IRT, CDM aims to provide information about the attributes or skills that are finer-grained and more relevant to classroom instruction and learning. As a state-of-the-art psychometric methodology, CDM is not typically offered as a regular course in most graduate measurement programs. By giving an overview of CDM, this workshop will be useful for faculty and students with a particular focus on educational measurement, as well as those interested in expanding their psychometric repertoire.

Furthermore, at present, only a few assessments have been developed using the CDM framework, one of which is the PR test developed by the lead instructor of this workshop, Dr. de la Torre, based on his NSF grant. He will share his experience in developing the assessment, which could be useful for participants interested in developing their own diagnostic tests.

Last, there are only a few computer programs for CDM analyses at present, and many of them suffer from various limitations. In this workshop, the GDINA R package developed by the instructors will be introduced. This package overcomes several drawbacks in existing software packages, and offers a wide range of functions for CDM analyses, such as calibration of various diagnostic models, validation of the Q-matrix, and detection of differential item functioning. After this workshop, participants are expected to be able to conduct various CDM analyses on their own.

Workshop Schedule

The workshop is structured to include nine sessions ranging from 45 to 50 minutes that are carefully designed to cover various aspects of CDMs, with an emphasis on both conceptual understanding and practical application.

Of particular note, seven out of the nine sessions will include practical hands-on activities using the GDINA R package, a powerful and flexible tool for conducting CDM analysis. Participants will have the opportunity to engage directly with the software. This blend of theory and practice is intended to enhance comprehension, providing participants with tangible skills they can apply in their future research. The structure of and additional details about the workshop are given below.

Welcome
9:00 AM – 9:10 AM
Session 1
Introduction to Diagnostic Assessments
9:10 AM – 10:00 AM
Session 1 consists of two parts. The first part introduces the diagnostic modeling framework in the context of educational assessment, and discusses the fundamental difference between traditional item response and CDMs. The second part introduces a diagnostic test for PR, and discusses its development, which includes attribute definition and Q-matrix development.
Session 2
The G-DINA Model Framework
10:00 AM – 10:45 AM
Session 2 covers the basic notation, terminology, input and output involved in CDM. Several widely used models such as the DINA, DINO, R-RUM, and LLM are introduced and then placed within the G-DINA model framework. Topics such as different joint distributions for attributes, estimation methods, and monotonicity constraints will also be discussed.
Break
10:45 AM – 11:00 AM
Session 3
Model Calibration in the GDINA R Package
11:00 AM – 11:45 AM
Session 3 focuses on the GDINA R package. Using this package, participants will learn how to estimate the G-DINA model and the reduced models it subsumes, how to model joint attribute distributions, how to impose monotonicity constraints when fitting these models, and how to obtain individual’s attribute estimates. Participants will calibrate the PR data and interpret the relevant information.
This session includes hands-on exercises
Session 4
Model Identifiability
11:45 AM – 12:30 AM
Session 4 discusses the importance of ensuring that diagnostic assessments designed to be analyzed using CDMs yield unique item parameter estimates and examinee classifications, as in, the model is identified. The concept of strict and generic identifiability for different types of CDMs will be discussed. Participants will be introduced to the R package cdmTools, which can be used in conjunction with the GDINA package, to assess the identifiability of the Q-matrices associated with particular diagnostic assessments.
This session includes hands-on exercises
Lunch Break
12:30 AM – 1:30 PM
Session 5
Model Fit Evaluation
1:30 PM – 2:20 PM
Session 5 addresses the issue of model appropriateness when fitting CDMs. Different item-level and test-level statistics provided as part of the GDINA package output will be discussed. In addition to numerical results, a graphical tool that can facilitate the evaluation of model-data fit will be covered.
This session includes hands-on exercises
Session 6
Model Comparison
2:20 PM – 3:10 PM
Session 6 covers test- and item-level statistical procedures for formally comparing saturated and reduced models under the G-DINA model framework. The theoretical and practical importance of model selection in a context where numerous, seemingly interchangeable models are available will be highlighted.
This session includes hands-on exercises
Session 7
Q-matrix Validation
3:10 PM – 3:55 PM
Session 7 introduces an efficient method for empirically validating provisional Q-matrices based on the G-DINA model framework. The method can be used in conjunction with all the reduced models subsumed by the G-DINA model. Participants will be given an opportunity to validate Q-matrices based on this method using the GDINA package.
This session includes hands-on exercises
Break
3:55 PM – 4:10 PM
Session 8
DIF and Classification Accuracy
4:10 PM – 5:00 PM
Session 8 discusses the criticality of test fairness in the CDM context. Various procedures for assessing differential item functioning in the G-DINA model framework will be covered. The session also covers the calculation of test-, item-, and attribute-level classification accuracy indices, which are indicators of test reliability and critical for applied researchers to assess the practical utility of a diagnostic test.
This session includes hands-on exercises
Session 9
Polytomous Attributes
5:00 PM – 5:50 PM
Session 9 introduces the pG-DINA model, a generalization of the G-DINA model designed to handle polytomous attributes. Such a CDM is deemed more appropriate when finer-grained inferences beyond student mastery/nonmastery are of interest (e.g., no, basic, or advanced mastery). Its current implementation in the GDINA package will be discussed.
This session includes hands-on exercises
Wrap-up
5:50 PM – 6:00 PM

About the Facilitators

Jimmy de la Torre
The University of Hong Kong

Prof. Jimmy de la Torre is Professor at the Faculty of Education at The University of Hong Kong. His research interests include latent variable models for educational and psychological measurement, and the use of assessment to improve classroom instruction and learning. Jimmy was a recipient of the 2008 Presidential Early Career Award for Scientists and Engineers given by the White House, and the 2009 Jason Millman Promising Measurement Scholar Award given by the National Council on Measurement in Education. He is the President of the Psychometric Society for the 2026-2027 term and a fellow of the American Educational Research Association.

Wenchao Ma
University of Minnesota

Prof. Wenchao Ma is an Associate Professor in the Department of Educational Psychology at the University of Minnesota and the John P. Yackel Professor in Educational Assessment and Measurement. His research interests lie in educational and psychological measurement in general, and cognitive diagnosis modeling and item response theory in particular. Wenchao was a recipient of the 2021 Jason Millman Promising Measurement Scholar Award and the 2017 Bradley Hanson Award given by the National Council on Measurement in Education and the 2018 Outstanding Dissertation Award given by the American Educational Research Association. He is the lead developer of the GDINA R package that will be used in this workshop.

Sangbeak Ye
Florida Atlantic University

Prof. Sangbeak Ye is Associate Professor of Research Methodology in the Department of Educational Leadership and Research Methodology at Florida Atlantic University. Trained as a statistician, his research program integrates psychometrics, sequential analysis, and Bayesian methods to advance adaptive learning and intelligent tutoring systems. He specializes in latent class modelling, especially cognitive diagnosis modelling, with applications ranging from educational measurement to workforce training and marketing analytics. Prior to joining Florida Atlantic University, Sangbeak was at the Methods Center of the University of Tübingen and the University of Missouri–Kansas City.

For additional information, contact:

📧 (for general enquiries concerning the workshop)
📧 edfacor@hku.hk (for enquiries concerning registration)
📍 Faculty of Education, The University of Hong Kong

About the speaker(s):
Register Here
 linkedin facebook pinterest youtube rss twitter instagram facebook-blank rss-blank linkedin-blank pinterest youtube twitter instagram