a-stratified multistage computerized adaptive testing

Computerized adaptive tests (CAT) commonly use item selection methods that select the item which provides maximum information at an examinee's estimat ed trait level, However, these methods can yield extremely skewed item expo sure distributions. For tests based on the three-parameter logistic model, it was found that administering items with low discrimination parameter (a) values early in the test and administering those with high a values later was advantageous; the skewness of item exposure distributions was reduced w hile efficiency was maintained in trait level estimation. Thus, a new multi stage adaptive testing approach is proposed that factors a into the item se lection process. In this approach, the items in the item bank are stratifie d into a number of levels based on their a values. The early stages of a te st use items with lower us and later stages use items with higher us. At ea ch stage, items are selected according to an optimization criterion from th e corresponding level. Simulation studies were performed to compare a-strat ified CATs with CATs based on the Sympson-Hetter method for controlling ite m exposure. Results indicated that this new strategy led to tests that were well-balanced, with respect to item exposure, and efficient. The a-stratif ied CATs achieved a lower average exposure rate than CATs based on Bayesian or information-based item selection and the Sympson-Hetter method.