Is Bayesian Estimation Proper for Estimating the Individual's Ability?
Abstract
There is a widespread belief among psychologists in the area of applied measurement that Bayesian estimation is better than the maximum likelihood estimation because of the additional information, i.e., prior. For example, many researchers in the computerized adaptive testing use Bayesian methods in the estimation of the examinee's ability. In this paper, this myth is debated theoretically, and in relation with the behavioral reality. Simulation studies are also used to show how biases caused by priors will affect the resultant estimation of the examinee's ability. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 15, 1980
- Accession Number
- ADA087829
Entities
People
- Fumiko Samejima
Organizations
- University of Tennessee