Applicability of neural test theory

–Discussing methodological problems and applying a polytomous model–

Abstract

Neural test theory (NTT) is a data analysis method that has gradually become popular in educational measurement and psychometrics. In NTT, examinees are clustered to a discrete latent rank. The author noted several technical issues for future researches of NTT, such as accuracy of estimates for latent ranks and ICRP, consistency of estimates for latent ranks over item sampling, comparison among several optimizing criteria, and construction and improvement of algorithm for NTT models. In the present research, the author performed a simulation study by using polytomous NTT for ordered data, to compare consistency of estimates for latent ranks over between NTT and another method using total test score. Finally, a real data example for essay test data was shown by using polytomous NTT, and the author compared these results with methods based on item response theory and total test score.