Final Report for Contract N00014-85-K-0113 (University of Missouri)
Abstract
The overall goal of this project is to develop new Bayesian procedures for mental testing. A typical test, which is studied here, consists of k test items administered to n examinees. The data consists of an nxk matrix of binary responses indicating which of the k items are scored correctly and which incorrectly by each of the n examinees. The statistical procedures are based on the assumption that there is a model which specifies the probability of a correct response to each item as a function of an unidimensional ability. Such functions are assumed to belong to certain families such as, the two-parameter logistic (2PL) or three-parameter (3PL) curves. These curves are identified by parameters called item parameters. When these models are used for testing, a set of items is initially calibrated using a moderately large value for n (the sample size). The calibration consists of estimating the item parameters. The calibrated curves are then used to score abilities of new examinees. Keywords: Bayer item response theory.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1988
- Accession Number
- ADA199435
Entities
People
- Robert K. Tsutakawa