Report
Model Selection for Equating Testlet-Based Tests in the NEAT Design: An Empirical Study
2012
Abstract
For those tests solely composed of testlets, local item independency assumption tends to be violated. This study, by using empirical data from a large-scale state assessment program, was interested in investigates the effects of using different models on equating results under the non-equivalent group anchor-test (NEAT) design. Specifically, the primary purpose of this study was to apply the IRT true-score equating method to equating testlet-based tests using both testlet theory (TRT) model and bi-factor model. In addition, the equating results from using the TRT and bi-factor models were compared with those from using conventional dichotomous item response theory (IRT) models. The candidate models considered in this study included a series of conventional dichotomous IRT models, Testlet model, and bi-factor model. The results echoed with those in Lee et al. (2001) in that equating using models that can account for item dependency in general tend to yield closer equating relationship to the traditional equating methods than the conventional IRT models. Limitations and further studies were also discussed. (Contains 4 figures and 8 tables.)
Details
- Title: Subtitle
- Model Selection for Equating Testlet-Based Tests in the NEAT Design: An Empirical Study
- Creators
- Wei HeFeifei LiEdward W WolfeXia Mao
- Resource Type
- Report
- Number of pages
- 22 pages
- Language
- English
- Date published
- 2012
- Academic Unit
- Psychological and Quantitative Foundations
- Record Identifier
- 9985134749402771
Metrics
1 Record Views