Criterion-Referenced Testing: A Critical Analysis of Selected Models
Abstract
Several mathematical models for use in criterion-referenced testing are reviewed and compared. The models are evaluated on both formal-analytic and empirical grounds. Predictive models include probabilistic formulations, a binomial model, and a Bayesian model. Descriptive methods include a categorization scheme, a one-parameter logistic model, and linear regression. An empirical method for relating mastery criteria to derived educational outcomes is also included. Problems inherent in each model or class of models are described. Such problems include tenability of assumptions, ease of application, assessment of item characteristics, and assessment of the model's fit to data. Each method/model appears to be appropriate for specific types of testing situations, although further development will depend upon computer simulation and empirical research.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1978
- Accession Number
- ADA061569
Entities
People
- Angelo Mirabella
- Frederick H. Steinheiser Jr.
- George B. Macready
- Kenneth I. Epstein
Organizations
- University of Maryland