What it’s about

In this paper, we examine whether PISA creative-thinking ratings can reasonably be scaled with the generalized partial credit model when the scoring instructions require different attributes at different thresholds.

Abstract

Using the PISA 2022 Creative Thinking data, this paper tests the assumption that a single latent dimension underlies the progression from no credit to partial and full credit. We show that the rating instructions instead imply within-rating multidimensionality, because appropriateness and originality or diversity are invoked differently across thresholds. Generalized IRTree models are used to represent this sequential scoring process, and they outperform unidimensional alternatives. The findings clarify the psychometric consequences of PISA’s rating logic and support a more explicit modeling of the distinct attributes involved in creative-thinking judgments.

How to access the paper

Go to the journal page for this article here.

Download the PDF here.