Özet:

The objective of this study is to compare the multi-pointed classified reaction model with the theory of substance reaction models for substances scored through binary coding in terms of model-data compatibility, accuracy of capacity cutting and test information functions. The study group consists of 1,500 students selected from the students who participated in the 2012 Level Assessment Exam. The statistical analysis in accordance with the purpose of the research was carried out using R studio software, Multilog, Lisrel 8.80 and SPSS package programs. According to the results of the analyses aimed at testing the conformity of the model, on the basis of the general conformity and substance conformity, STM, 2PLM and 3PLM showed adequate conformity to the research data, and the highest conformity among these models was observed as 3PLM. In terms of capacity parameters, it has been found that STM has resulted in more accurate capacity cuts than MTK models scored through binary coding. Considering the test information functions, it was observed that in the low-capacity ranking, STM provided more information than all the MTK models scored through binary coding. However, although the capacity cuts based on STM were made more accurate, the test found that under this model, less information on capacity cuts was provided compared to 3PLM. However, it is believed that the STM will also benefit from testing and development of all aspects of the substances by examining the effectiveness of scratchers for already selected testing substances. In this regard, it is expected that the STM will contribute to the work of creating a material bank in particular in the application.

Özet:

The purpose of this study is to examine the dichotomous item response theory models and nominal response model in terms of model-data fit, accuracy of ability estimation, and test information functions in a comparative way. The study group was composed of 1500 students taken 2012-SBS. For the research purposes, the analysis were conducted through the R Studio software, Multilog, Lisrel 8.80 and SPSS. In terms of model data fit, it was observed that nominal response model, 2 parameter logistic model and 3 parameter logistic model fits the data both in terms of overall fit and item fit, and 3PLM is the best fitting model among them. The abilities were most precisely estimated with NRM . Lastly, in terms of test information functions, NRM provided the most information than the dichotomus IRT models at the low ability range. However, NRM provided less information in terms of total test information function than 3PLM, although the ability parameters were estimated with better accuracy with NRM than 3PLM. However, it is considered that nominal response model would be beneficial to be improved of items with all aspcets by examining the effectiveness of distractors. Therefore, it is believed that nominal response model, especially in practice, would contribute to studies aimed at constructing item bank.

Anahtar Kelimeler: