Optimal Item Selection in Multidimensional Adaptive Testing

• Activity Number: 193
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Education
• Author(s): Haskell Sie*+ and James L Rosenberger
• Companies: Penn State University and Penn State University
• Keywords: Item selection ; Optimal design ; Adaptive tests

Abstract

In the recent years, Multidimensional Computerized Adaptive Testing has become more popular due to increase interest in obtaining as much information as possible about examinees' various abilities that affect their responses on the test. Crucial in the administration of such tests is the choice of item selection methods that will help to minimize the uncertainty in estimating examinees' latent abilities. Using the two-dimensional three-parameter logistic item response model, we will extend the results of Mulder & van der Linden (2009) by checking the accuracy of ability estimation when test items are selected based on the Kullback-Leibler information index (KLII), simplified KL information index (SKLII), and mutual information (MI). A simulation study assesses each of the methods in three cases: where all ability dimensions are of interest, where only some of the ability dimensions are of interest, and where interest lies in a certain linear combination of the multidimensional abilities. The characteristic of item parameters preferred by each method is also assessed. In addition, item overlap rates are calculated to show similarities between the methods.