This project aims to develop an automated method for obtaining a valid and reliable instrument using the Rasch model as a basis. Using this automated method, we aim to eliminate the drawbacks of manual Rasch analysis. The first drawback is that Rasch analysis involves different kinds of statistical tests, but there are no strict rules about which test should be performed first. As a result, the decision on how to conduct the analysis depends on one's knowledge and experience, and different decisions may result in different final instruments. Secondly, the Rasch analysis validates instruments by removing misfits sequentially. Henceforth, the analysis becomes more burdensome when dealing with a large number of items. Thus, we introduce a novel criterion to judge the instrument's quality: the higher the score, the better the instrument. With this criterion, we aim for the highest-scoring instrument. We, however, are careful enough to call this a semi-automated method rather than a fully automated one since we are aware that experts' knowledge is invaluable. As a final point, the datasets presented here are used in empirical studies. The simulated datasets are generated randomly and independently to illustrate the characteristics of the semi-automated method. Using real-world datasets, we show that the semi-automated method produces comparable results to the manual Rasch method.

These real-world datasets consist of preprocessed versions of three publicly available datasets that can be found in the publications' supplementary data of Vijaya K. Gothwal et al. , Shu Imaizumi and Yoshihiko Tanno , and Elisa Morrone et al. .