You are here:

The D-Optimality Item Selection Criterion in the Early Stage of CAT: A Study with the Graded Response Model
ARTICLE

, ,

Journal of Educational and Behavioral Statistics Volume 33, Number 1, ISSN 1076-9986

Abstract

During the early stage of computerized adaptive testing (CAT), item selection criteria based on Fisher"s information often produce less stable latent trait estimates than the Kullback-Leibler global information criterion. Robustness against early stage instability has been reported for the D-optimality criterion in a polytomous CAT with the Nominal Response Model and is shown herein to be reproducible for the Graded Response Model. For comparative purposes, the A-optimality and the global information criteria are also applied. Their item selection is investigated as a function of test progression and item bank composition. The results indicate how the selection of specific item parameters underlies the criteria performances evaluated via accuracy and precision of estimation. In addition, the criteria item exposure rates are compared, without the use of any exposure controlling measure. On the account of stability, precision, accuracy, numerical simplicity, and less evidently, item exposure rate, the D-optimality criterion can be recommended for CAT. (Contains 9 figures and 1 table.)

Citation

Passos, V.L., Berger, M.P.F. & Tan, F.E.S. (2008). The D-Optimality Item Selection Criterion in the Early Stage of CAT: A Study with the Graded Response Model. Journal of Educational and Behavioral Statistics, 33(1), 88-110. Retrieved January 22, 2020 from .

This record was imported from ERIC on April 18, 2013. [Original Record]

ERIC is sponsored by the Institute of Education Sciences (IES) of the U.S. Department of Education.

Copyright for this record is held by the content creator. For more details see ERIC's copyright policy.

Keywords