In examinee-selected-item (ESI) design, examinees are requested to select a fixed number of items from a given set of items to respond. This design may enhance students’ learning motivation and reduce students’ testing anxiety. However, responses to not-selected items are often (if not always) missing not at random (MNAR). In this study, we proposed non-ignorable missing examinee-selected item model (NMESIM) that directly quantifies the non-ignorability. The parameter recovery and the consequences of model misspecification were evaluated using simulations. An empirical example was given to demonstrate the implication and applications of the NMESIM. Given these promising results, we are confident to recommend the NMESIM for the analysis of examinee-selected items, such that the attractions of the ESI design can be maintained.
|Publication status||Published - Apr 2014|