Workshop 8
Item Response Theory: Introduction to concepts, models, parameter estimation and fit, and several applications
Ronald K. Hambleton
University of Massachusetts at Amherst, USA
Abstract
Many testing agencies and researchers would like to use item response theory (IRT) models for developing, scoring, identifying bias, and equating of their aptitude, achievement, and personality tests. These IRT models, too, can be used to provide the measurement underpinnings for new test designs such as multi-stage testing and computer-adaptive testing. In this workshop, we will survey the following topics and provide several examples and practical experiences:
- Shortcomings of classical test theory that have inspired the development of IRT models, and basic classical test theoretic concepts such as reliability and item analysis,
- Specific IRT models for fitting binary and polytomously-scored data (e.g., 1-, 2-, and 3-parameter logistic models, graded response model),
- Basics of item and ability parameter estimation,
- Graphical and statistical approaches for assessing model fit (e.g., RESID PLOTS-2),
- Introduction to IRT software (e.g., BILOG-MG, PARSCALE),
- Development of tests using item and test and target information functions, and relative efficiency,
- Computer-based testing: Issues, designs, item exposure, and advantages and disadvantages,
- Identification of potentially biased test items due to culture, content, translation, and other factors,
- Follow-up readings and research.
|