An Investigation of One and Three Parameter Item Response Models in a Field Setting, with Implications for Computerized Adaptive Testing
Abstract
The general objective of computer driver adaptive testing is to accurately estimate an individual's position on the underlying trait the test purports to measure. Conventional paper-and-pencil test administration typically suffers from several sources of error in the measurement of an individual's ability. Conventional peaked tests are designed to discriminate most effectively at a single ability level and thus assume that most individuals taking the test fall into this category. The results of this lower precision of measurement is lower overall reliability, and lower validity as well. Two recent developments have enabled the psychometrician to more accurately assess the status of individuals on measurable traits: adaptive testing and latent trait test theory. Adaptive testing enables adapting each test to fit characteristics of each individual tested.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1981
- Accession Number
- ADA214812
Entities
People
- Bronel R. Whelchel
Organizations
- Morgan State University