Using response times to detect examinees with item pre-knowledge in a testlet response model

Sheng Yun HUANG, Wen Chung WANG

Research output: Contribution to conferencePapers

Abstract

Review Articles Background: In recent years, increased attention has been given to response times in item response models. Several studies have noted that response times provide an excellent piece of information in identifying examinees with item pre-knowledge. Standard item response models, such as the 3-parameter logistic model, have been extended to incorporate response times (Wang and Hanson, 2005). Review Articles Aims: As testlet-based items have been widely used and testlet response models have been developed, it is natural to extend tetslet response models to incorporate response times and to examine the effect of item pre-knowledge detection, which is the major purpose of this study. Review Articles Method: We first formulated such a testlet response model with response times and conducted a series of stimulations to assess item pre-knowledge detection under testlet response model. The computer program FORTRAN 90 was used to conduct the simulation. Review Articles RASCH: The Rasch testlet model with response times was formulated. Review Articles Results: The results show that response times were very useful in item pre-knowledge detection, especially for persons with moderate ability levels. Review Articles Conclusions: Response times are a piece of useful information for detecting examines with item pre-knowledge when testlet-based items are administered.
Original languageEnglish
Publication statusPublished - Jul 2009

Citation

Huang, S.-Y., & Wang, W.-C. (2009, July). Using response times to detect examinees with item pre-knowledge in a testlet response model. Paper presented at the Pacific Rim Objective Measurement Symposium 2009 (PROMS 2009) Hong Kong, The Hong Kong Institute of Education, China.

Fingerprint Dive into the research topics of 'Using response times to detect examinees with item pre-knowledge in a testlet response model'. Together they form a unique fingerprint.