Modelling the Differences in Social and Emotional Skills with Polytomous Explanatory IRT: The Example of Assertiveness Skill

Abstract

Explanatory item response theory models can simultaneously decompose the covariance between persons and items, as well as analyze items by adding item-related predictors for differences between item difficulties and/or person-related predictors for differences between individuals. In the current study, we calculated the parameter estimations regarding the skill of assertiveness according to the rating scale model and partial credit model, which are descriptive (traditional) item response theory models as well as latent regression partial credit model including only person-level predictors, and then examined the results comparatively. We used the raw score belonging to the skill of assertiveness of Türkiye belonging to the OECD Social and Emotional Skills Study, and we included gender, socioeconomic level, perceived relationships with teachers, bullying at school, sense of belonging at school, global mindedness, and test anxiety as person-level predictors. Current study findings suggests that; (1) the latent regression partial credit model produces a better data fit when compared to the rating scale model and partial credit model, and (2) sense of belonging at school, global mindedness, and socioeconomic level are significant predictors to explain the differences between persons. We discussed the current study findings in terms of the rich body of knowledge provided by explanatory item response theory and presented some suggestions.