Title: Item calibration methods for multi-stage design
Authors: Chun Wang - University of Minnesota (United States) [presenting]
Abstract: Many large scale educational surveys have moved from linear form design to multistage testing (MST) design. A MST tailors the set of items (i.e., target block) a student sees to the students individual ability level, so that no examinee receives too many overly easy or difficult items. Consequently, MST can provide more accurate latent trait estimates ($\theta$) using fewer items than required by linear tests. However, MST generates incomplete response data by design; hence questions remain as to whether the item calibration procedure for the traditional linear form can still apply? If not, what adjustments should be made? Deriving from the missing data mechanism, two new item calibration methods will be proposed, namely the multiple-group with truncated normal prior and the multiple-group with empirical histogram prior. They will be compared to the traditional single-group marginal maximum likelihood (MMLE) and multiple-group MMLE in terms of item parameter recovery.