Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Sufficient and Necessary Conditions for the Identifiability of DINA Models with Polytomous Responses

Mengqi Lin, Gongjun Xu

TL;DR

This work addresses identifiability for CDMs with polytomous responses by developing a generalized $T$-matrix framework and deriving explicit necessary and sufficient conditions. For GPDINA, identifiability is achieved under conditions $\text{C1-C3}$ on a complete $\mathbf Q$-matrix, including each attribute appearing in at least three items and distinct core columns, with a one-to-one mapping between $P(R|\cdots)$ and the attribute distribution. For Sequential DINA, identifiability is established via the $T^s$-matrix under conditions $\text{S1-S3}$ on the first-categories matrix $\mathbf Q^1$, with discussion of the necessity of these conditions and potential relaxations using information from higher categories. The paper illustrates how these theoretical results translate into practical guidelines for test design and highlights identifiability challenges in real data (PISA TIMSS), while noting possible extensions to DINO and generic identifiability concepts. Overall, the results extend identifiability theory from binary to polytomous CDMs and provide actionable criteria for ensuring estimable parameters in polytomous cognitive diagnostics.

Abstract

Cognitive Diagnosis Models (CDMs) provide a powerful statistical and psychometric tool for researchers and practitioners to learn fine-grained diagnostic information about respondents' latent attributes. There has been a growing interest in the use of CDMs for polytomous response data, as more and more items with multiple response options become widely used. Similar to many latent variable models, the identifiability of CDMs is critical for accurate parameter estimation and valid statistical inference. However, the existing identifiability results are primarily focused on binary response models and have not adequately addressed the identifiability of CDMs with polytomous responses. This paper addresses this gap by presenting sufficient and necessary conditions for the identifiability of the widely used DINA model with polytomous responses, with the aim to provide a comprehensive understanding of the identifiability of CDMs with polytomous responses and to inform future research in this field.

Sufficient and Necessary Conditions for the Identifiability of DINA Models with Polytomous Responses

TL;DR

This work addresses identifiability for CDMs with polytomous responses by developing a generalized -matrix framework and deriving explicit necessary and sufficient conditions. For GPDINA, identifiability is achieved under conditions on a complete -matrix, including each attribute appearing in at least three items and distinct core columns, with a one-to-one mapping between and the attribute distribution. For Sequential DINA, identifiability is established via the -matrix under conditions on the first-categories matrix , with discussion of the necessity of these conditions and potential relaxations using information from higher categories. The paper illustrates how these theoretical results translate into practical guidelines for test design and highlights identifiability challenges in real data (PISA TIMSS), while noting possible extensions to DINO and generic identifiability concepts. Overall, the results extend identifiability theory from binary to polytomous CDMs and provide actionable criteria for ensuring estimable parameters in polytomous cognitive diagnostics.

Abstract

Cognitive Diagnosis Models (CDMs) provide a powerful statistical and psychometric tool for researchers and practitioners to learn fine-grained diagnostic information about respondents' latent attributes. There has been a growing interest in the use of CDMs for polytomous response data, as more and more items with multiple response options become widely used. Similar to many latent variable models, the identifiability of CDMs is critical for accurate parameter estimation and valid statistical inference. However, the existing identifiability results are primarily focused on binary response models and have not adequately addressed the identifiability of CDMs with polytomous responses. This paper addresses this gap by presenting sufficient and necessary conditions for the identifiability of the widely used DINA model with polytomous responses, with the aim to provide a comprehensive understanding of the identifiability of CDMs with polytomous responses and to inform future research in this field.
Paper Structure (34 sections, 15 theorems, 129 equations, 4 tables)

This paper contains 34 sections, 15 theorems, 129 equations, 4 tables.

Table of Contents

  1. Introduction
  2. Model Setup
  3. The GPDINA model
  4. Identifiability.
  5. The Seqential DINA model
  6. Identifiability.
  7. Relationship between GPDINA and Sequential DINA models
  8. Identifiability
  9. Generalized $\mathbf T$-matrix for CDMs with polytomous responses
  10. $\mathbf T$-matrix for GPDINA
  11. $\mathbf T$-matrix for Sequential DINA model
  12. Identifiability of the GPDINA model
  13. Identifiability of the Sequential DINA model
  14. Data Examples
  15. Identifiability of the GPDINA model: a PISA 2000 data example.
  16. ...and 19 more sections

Key Result

Proposition 1

Let $\Delta = \{j \in [J]: \mathbf q_j = \mathbf 0\}$ denote the set of items whose $\mathbf q$-vectors are zero, then the GPDINA model parameters with $\mathbf Q$-matrix are identifiable if and only if the GPDINA model parameters with $\mathbf Q_{-\Delta}$-matrix are identifiable, where $\mathbf Q_

Theorems & Definitions (44)

  • Example 1
  • Proposition 1
  • Example 2
  • Proposition 2
  • Lemma 1
  • Example 3
  • Lemma 2
  • Example 4
  • Theorem 1
  • Remark 1
  • ...and 34 more