Özet:

Cognitive diagnostic models enable the creation of individuals' attribute profiles through particular categorical latent variables. In order to collect evidence of whether the inferences drawn from these models are valid or not, it is of great importance to reveal the model-data fit and to identify the models that fit. This study aims to determine whether reduced DINA and DINO models can be used instead of the saturated GDINA model determined within cognitive diagnostic modelling framework by using Wald statistics for each item level and to specify the model that best fits the data by comparing it at the item and test level. The data set of the study was obtained from 478 students and the answer pattern of the Symbolic Number Comparison test comprising 36 items. Analyses were performed in R 3.5.3 program and the GDINA package in the R program was used. In the first stage, the saturated GDINA model was taken as the basis, and the absolute and relative fit indices were evaluated together and it was seen that the saturated model gave the best fit. Afterward, using Wald statistics, instead of the saturated GDINA model, the availability of reduced DINA and DINO models for each item-level with no significant loss in model-data fit was examined.

Anahtar Kelimeler: