Implementation of Cognitive Diagnosis Modeling using the GDINA R Package

Abstract

Purpose: Well-designed assessment methodologies and various cognitive diagnosis models (CDMs) to extract diagnostic information about examinees’ individual strengths and weaknesses have been developed. Due to this novelty, as well as educational specialists’ lack of familiarity with CDMs, their applications are not widespread. This article aims at presenting the fundamentals of CDM and demonstrating the various implementations using a freeware R package, namely, the GDINA. Present article explains the basics of CDM and provide sufficient details on the implementations so that it may guide novice researchers in CDM applications.

Research Methods: The manuscript starts with presenting the CDM terminology, including input and output of a CDM analysis. The introduction section is followed by generalized deterministic noisy and gate model framework. A brief description of the package GDINA is also provided. Then, numerical examples on various CDM analyses are provided using the R package with a graphical user interface. The paper is concluded by some additional functions and concluding remarks. Results and Implications for Research and Practice: Although other software programs are also available, using the GDINA package offers users some flexibilities such as allowing estimation of a wide range of CDMs and allowing nonprogrammers to benefit from this package through the GUI. In addition to ordinary CDM analyses, GDINA package further allows users to apply model selection at the test- and item-level to make sure that the most appropriate CDM (i.e., CDM that best explains the attribute interactions in the item) is fitted to the response data. Furthermore, to identify possible item-attribute specification mistakes in the Q-matrix, implementation of an empirical Q-matrix validation method is available in the GDINA package. Lastly, this package offers various handy graphs, which can be very useful in emphasizing important information and comparing various parameters and/or statistics.