NotesFAQContact Us
Search Tips
ERIC Number: ED550357
Record Type: Non-Journal
Publication Date: 2012
Pages: 194
Abstractor: As Provided
Reference Count: N/A
ISBN: 978-1-2677-9579-3
Structural Equation Model Approach to the Use of Response Times for Improving Estimation in Item Response Models
Sen, Rohini
ProQuest LLC, Ph.D. Dissertation, University of Connecticut
In the last five decades, research on the uses of response time has extended into the field of psychometrics (Schnikpe & Scrams, 1999; van der Linden, 2006; van der Linden, 2007), where interest has centered around the usefulness of response time information in item calibration and person measurement within an item response theory. framework. Motivated by van der Linden's 2007 hierarchical modeling of response and response times, this study proposes the introduction of a new model within a structural equation framework (SEM-IRTRT model) that incorporates response and response times to improve item parameter and ability estimation. The model uses the well-established single factor confirmatory factor analysis to model a two-parameter IRT model as well as a single factor confirmatory factor analysis model parameterization of van der Linden's 2006 lognormal RT model (Finger & Chee, 2009) where the impact of the latent ability parameter on latent person speed is evaluated by defining a direct effect relationship from speed to ability. These models were employed to explore the impact of using response times as collateral information on accuracy of IRT item and person parameters. Two simulations studies were carried out to (1) establish equivalence of the proposed model with the established van der Linden (2007) model for dichotomous items and (2) to extend study 1 to items with polytomous responses. Overall, in both studies 1 and 2, it was found that for IRT item parameters (discriminations and difficulties), collateral information provided by response time had little impact on accuracy of estimation of these parameters. However, using response time in estimation did improve recovery of examinee abilities for the IRTRT models. This has important implications for future use of SEM IRTRT models when using RTs as collateral information especially for items with polytomous responses. Furthermore, for the study with polytomous items, it was found that SEM-IRTRT model performs best in recovering item and person parameters for smaller test lengths and sample sizes, which is beneficial for test design especially in terms of minimizing test construction related costs. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page:]
ProQuest LLC. 789 East Eisenhower Parkway, P.O. Box 1346, Ann Arbor, MI 48106. Tel: 800-521-0600; Web site:
Publication Type: Dissertations/Theses - Doctoral Dissertations
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A