A quick item selection method in computerized adaptive testing for ranking items

Ranking items have been widely used in non-cognitive tests such as personality tests and career interest tests. The Rasch model for ipsative tests with multidimensional pairwise comparison items was recently developed and their corresponding CAT algorithms were investigated (Chen & Wang, 2013, 2014). Moreover, this model has been extended to account for ranking items in which more than two statements are to be ranked (Qiu & Wang, 2015). To facilitate this new model for ranking items in a CAT environment, we investigated how a ranking item is selected in a reasonably short time. As this model tends to be high-dimensional and there are a huge number of possible ranking items in an item bank, standard item selection methods that are based on Fisher item information matrix become infeasible in real time. We proposed a quick-and-dirty method for item selection and evaluate its performance with simulations. The result showed this method could select a ranking item in a short time without sacrificing too much efficiency. In the future, we will further implement control procedures for item exposure.