New Item Response Theory Models for Rater Errors

​

Item response theory (IRT) and Rasch measurement has contributed significantly to the assessment of rater errors. Historically, rater errors of severity and inconsistency are the two major foci; leaving centrality and similarity almost untouched. Given they are common in rater data, failing to consider such errors makes test scores unfair and subsequent decisions erroneous (e.g., college admission). This project aims to overcome this by inventing new IRT models in both frameworks that consider not only severity and inconsistency but also centrality and similarity. It will involve simulation and empirical studies to evaluate the new models and demonstrate their applications. Computer source codes and examples will be made available to researchers and practitioners. Pilot studies will be conducted to extend the new models and create new research lines for future studies.

​