An investigation of the prevalence and difficulty of reading comprehension's sub-skills by the G-DINA model

Reading comprehension is one of the most important skills of English language, specifically in academic settings. This skill has been investigated time and again from different perspectives, of which educational measurement is the focus of the present research. This study aims at defining these underlying sub-skills, examining their prevalence and difficulty, and estimating the difficulty of the items of a large-scale exam. The G-DINA model (Ma & de la Torre, 2017), which is a cognitive diagnostic model, was selected as the statistical method of data analysis. To this end, the subtest of reading from National University Entrance Exam, master's level, was selected. The underlying sub-skills of the test were extracted through three main sources of concurrent literature, students' think-aloud protocols, and expert panel's judgment. The extracted sub-skills along with the students' scored responses were used as the input for the GDINA package in R programming software. Four sub-skills were defined for the test and the outputs related to attribute/sub-skill prevalence, sub-skill difficulty, and item difficulty were reported in CDM framework. In the end, the probable reasons for the obtained outputs were discussed in the context of reading comprehension