Application of Explanatory Item Response Theory Models in Psychometric Analysis of Items on E-learning Platforms

The thesis deals with Explanatory Items Response Theory models and describes their possible application on e-learning platforms in order to explain item difficulty and investigate its sources. The model of items in Russian language from the Yandex.Educational educational platform for 2-4 grade students is built using LLTM+e IRT model framework. The modeling is based on the result of experts’ evaluation of item content characteristics, metasubject skills required for item solving, item response format and its interactive realization, and other item features. The result showed that the Explanatory IRT models allow for explaining the large portion of items’ difficulty variance (in this paper – up to 96%) and estimation of difficulty sources. It proves possibility of separation of the items’ content effects and user interface characteristics, which was the purpose of the study. The results also suggest that item difficulty depends on user interface characteristics controlling for the item content effects. The results of Explanatory Items Response Theory modelling are useful for test-developers since such models can validate items and ideas on cognitive processing of them, provide insights on content value for e-learning platforms and predict the difficulty of new items. The results state that application of Explanatory IRT models on e-learning platforms provides opportunities to explain the observed behavior of respondents and make data analysis deeper.