Modeling missing-data processes: An IRTree-based approach

In large-scale assessments two missing responses can be discerned: “not-reached” and “skipped” responses . Both omissions may be related to the test takers’s proficiency, resulting in non-ignorable missingness. An IRTtree-based model with separate latent processes for “not-reached” and “skipped” responses is proposed, and illustrated using empirical data.status: publishe