Performed a simulation study to demonstrate the dramatic impact of capitalization on estimation errors on ability estimation in adaptive testing. Discusses four different strategies to minimize the likelihood of capitalization in computerized adaptive testing. (SLD)

Test security has often been a problem in computerized adaptive testing (CAT) because the traditional wisdom of item selection overly exposes high discrimination items. The a-stratified (STR) design advocated by H. Chang and his collaborators, which uses items of less discrimination in earlier stages of testing, has been shown to be very…

In the a-stratified method, a popular and efficient item exposure control strategy proposed by H. Chang (H. Chang and Z. Ying, 1999; K. Hau and H. Chang, 2001) for computerized adaptive testing (CAT), the item pool and item selection process has usually been divided into four strata and the corresponding four stages. In a series of simulation…

Compared five item selection rules with respect to the efficiency and precision of trait (theta) estimation at the early stages of computerized adaptive testing (CAT). The Fisher interval information, Fisher information with a posterior distribution, Kullback-Leibler information, and Kullback-Leibler information with a posterior distribution…

In the process of item calibration for a computerized adaptive test (CAT), many well-established calibrating packages show weakness in the estimation of item parameters. This paper introduces an on-line calibration algorithm based on the convexity of likelihood functions. This package consists of: (1) an algorithm that estimates examinee ability…

This paper suggests several item selection criteria for adaptive testing that are all based on the use of the true posterior. Some of the ability estimators produced by these criteria are discussed and empirically criticized. (SLD)

The average Kullback-Keibler (K-L) information index (H. Chang and Z. Ying, in press) is a newly proposed statistic in Computerized Adaptive Testing (CAT) item selection based on the global information function. The objectives of this study were to improve understanding of the K-L index with various parameters and to compare the performance of the…

In adaptive testing, item selection is sequentially optimized during the test. Since the optimization takes place over a pool of items calibrated with estimation error, capitalization on these errors is likely to occur. How serious the consequences of this phenomenon are depends not only on the distribution of the estimation errors in the pool or…

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…

When examinees are allowed to review responses on an adaptive test, can they "cheat" the adaptive algorithm in order to take an easier test and improve their performance? Theoretically, deliberately answering items incorrectly will lower the examinee ability estimate and easy test items will be administered. If review is then allowed,…

In order to estimate the classical coefficient of test reliability, parallel measurements are needed. H. Gulliksen's matched random subtests method, which is a graphical method for splitting a test into parallel test halves, has practical relevance because it maximizes the alpha coefficient as a lower bound of the classical test reliability…

A computerized adaptive test (CAT) administration usually requires a large supply of items with accurately estimated psychometric properties, such as item response theory (IRT) parameter estimates, to ensure the precision of examinee ability estimation. However, an estimated IRT model of a given item in any given pool does not always correctly…

Some alternative criteria for item selection in adaptive testing are proposed that take into account uncertainty in the ability estimates. A simulation study shows that the likelihood weighted information criterion is a good alternative to the maximum information criterion. Another good alternative uses a Bayesian expected a posteriori estimator.…

Effects on computerized adaptive test score of using a paper-and-pencil (P&P) calibration to select items and estimate scores were compared with effects of using computer calibration. Results with 2,999 Navy recruits support the use of item parameters calibrated from either P&P or computer administrations. (SLD)

This study investigated and compared the properties of five methods of item exposure control within the purview of estimating examinees' abilities in a computerized adaptive testing (CAT) context. Each of the exposure control algorithms was incorporated into the item selection procedure and the adaptive testing progressed based on the CAT design…