Comparison of Classification Accuracy and Parameters of DINA, DINO, HO-DINA and HO-DINO Models in the Framework of Cognitive Diagnosis in Health Education

Purpose: This study aims to compare the parameters of DINA, DINO, HO-DINA and HO-DINO models according to different sample sizes (500, 2000, 5000) and different item numbers (60, 120) based on the Q matrices created for different attributes in health education based on simulation data. Design/Methodology/Approach: In the simulation data, 50 replications were performed for each condition. In the study, two different Q-Matrixes were determined based on the learning domain determined by considering the 2018 TUS Spring Assessment Report and the taxonomy included in the Clinical assessment framework determined in Miller's 1990 study as the attributes dimension in the Q-Matrix in which matching of attribute and item is carried out. In the study, RMSEA, g and s parameters and classification accuracies were compared and under which conditions DINA, DINO, HO-DINA and HO-DINO models gave similar or different results were investigated. Findings: According to the research findings, the Q-Matrix, in which Fields levels were used as the attribute dimension, was the matrix that gave the best parameter results in all models. In addition, it has been determined that the models that give the best RMSEA, g and s parameters and classification accuracies are DINO and HO-DINO models in the analysis. Highlights: Based on the findings, when analyzing the results for the Basic Medical Sciences and Clinical Medical Sciences tests, it is evident that the Q matrix determined by Fields provides a better fit to the data, and moreover, it is advantageous for the Q matrix determined by Fields to be used for the TUS exam.