NotesFAQContact Us
Search Tips
ERIC Number: ED238940
Record Type: Non-Journal
Publication Date: 1983-Nov
Pages: 108
Abstractor: N/A
Reference Count: N/A
Optimal Measurement Considerations for Diagnostic Tests. Methodology Project.
Wilcox, Rand R.
This document presents a series of five papers describing issues in educational measurement. "A Simple Model for Diagnostic Testing When There Are Several Types of Misinformation" directly addresses the diagnostic issue. It describes a simple latent trait model for testing, examines use of erroneous algorithms, and illustrates the derivation of an optimal scoring rule for multiple choice test items. "Measuring Mental Abilities with Latent State Models" has three goals: to review the latent state models that have been proposed for measuring aptitude and achievement; to outline the measurement problems that can now be solved with latent state models; and to discuss how latent state and latent trait models are related. "Strong True Score Theory" reviews true score models in light of various assumptions about guessing. "Approximating Multivariate Distributions" suggests a simple approximation of multivariate distributions. The suggested method is compared with several other approximations. These comparisons indicate that the new approximation nearly always gives better results. "Unbiased Estimation in a Closed Sequential Testing Procedure" provides an optimal linear estimator of the proportion of items within an item domain that an examinee would answer correctly if every item were attempted. (Author/PN)
Publication Type: Collected Works - General; Reports - Research
Education Level: N/A
Audience: Researchers
Language: English
Sponsor: National Inst. of Education (ED), Washington, DC.
Authoring Institution: California Univ., Los Angeles. Center for the Study of Evaluation.