Impact of Including Item Response Time in a Model on Exam Score Rankings

• Activity Number: 423
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Huijuan Meng*+ and Jerry Gorham and Joy Matthews-Lopez
• Companies: Pearson, Inc. and Pearson, Inc. and National Association of Boards of Pharmacy
• Keywords: IRT model ; Rasch ; Item time ; Theta estimation ; Decision consistency

Abstract

Traditional item response theory (IRT) models in general involve estimating examinee latent trait(s) and item parameter(s) based on a matrix of item responses (binary or categorical). With computer-administered exams, though, the amount of time an examinee spends on each item becomes available. Researchers developed various models to fit item response time (RT) and to explore the advantages of adding this continuous variable in a model from different perspectives. In this paper, a hierarchical model (van der Linden, 2007) has been chosen to simultaneously account for item responses and times: on the first-level, a Rasch model is adopted for item response, whereas a lognormal model is selected for the RT; and the second level specified the associations between the parameters in the first level. The purpose of this paper is to investigate the impact, if any, of including RT in a model on exam scores. Decision consistency based on two sets of theta estimates under different models will be explored.