ERIC Number: ED454279
Record Type: Non-Journal
Publication Date: 2001-Apr
Reference Count: N/A
Nearest Neighbors, Simple Strata, and Probabilistic Parameters: An Empirical Comparison of Methods for Item Exposure Control in CATs.
Parshall, Cynthia G.; Kromrey, Jeffrey D.; Harmes, J. Christine; Sentovich, Christina
Computerized adaptive tests (CATs) are efficient because of their optimal item selection procedures that target maximally informative items at each estimated ability level. However, operational administration of these optimal CATs results in a relatively small subset of items given to examinees too often, while another portion of the item pool is almost unused. This situation both wastes a portion of the available items and can be a security risk. A number of exposure control methods have been developed to reduce this effect. In this study, the effectiveness of three methods was investigated in comparison to baseline conditions of No Control and Random item selection. These procedures were: (1) the Sympson-Hetter method (J. Sympson and R. Hetter, 1985); (2) the Nearest Neighbor method (R. Holmes and D. Segall, 1999); and (3) Stratified-a methods (H. Chang and Z. Ying, 1997). Using Monte Carlo procedures, these methods were examined under varying target maximum exposure rates. Results are reported in terms of pool usage, test precision and bias, both unconditionally and conditionally. Three methods were completely successful in preventing marginal administration rates beyond the specified target maximum, the Sympson Hetter and Nearest Neighbor methods and the Stratified-a method incorporating item freezing. (Contains 26 figures and 18 references.) (Author/SLD)
Publication Type: Numerical/Quantitative Data; Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Sponsor: Defense Manpower Data Center, Monterey, CA.
Authoring Institution: N/A