The Impact of Item Parameter Estimation on Computerized Adaptive Testing with Item Cloning (CT-02-06)
by Cees A. W. Glas, University of Twente, Enschede, The Netherlands
Item cloning techniques involve the application of an algorithm or algorithms to generate new test items based on the characteristics of existing test items. The term parent is sometimes used to describe the original items, and the term offspring is sometimes used to describe the new items that are generated by the cloning technique. The application of item cloning techniques can greatly reduce the cost of item writing and enhance the flexibility of item presentation. Another cost-saving factor of item cloning is that the item response theory (IRT) item statistics (commonly referred to as item parameters) that are calculated for each test item and used to select items for administration in a computerized adaptive test (CAT) are often assumed to be the same, or at least very similar, for parent and offspring items.
However, a potential negative consequence of cloning is the possible variation in the values of the item parameters for items cloned from the same parent. Recent research has shown that the presence of item parameter estimation error can lead to capitalization on these errors and hence to substantial loss of precision in the ability estimates derived for test takers.
This research presents a solution that accounts for the uncertainty in the item parameter estimates, thereby resolving the problem of capitalization on item parameter estimation errors. A simulation study is presented to illustrate how the method neutralizes the effect of capitalization on errors in item parameter estimates.