Accession Number : ADA112048

Title :   Latent Trait Model Contributions to Criterion-Referenced Testing Technology.

Descriptive Note : Final rept.,

Corporate Author : MASSACHUSETTS UNIV AMHERST LAB OF PSYCHOMETRIC AND EVALUATIVE RESEARCH

Personal Author(s) : Hambleton,Ronald K

PDF Url : ADA112048

Report Date : Feb 1982

Pagination or Media Count : 31

Abstract : The goals and results of an 18-month study addressing latent trait model applications to measurement problems arising in criterion-referenced testing are presented. The research studies described in the report cover the following areas: (a) problems with classical test models; introduction to latent trait models, features, assumptions, parameter estimation, and test and item information curves; building tests with latent trait models; (b) latent ability scales - uses, interpretations, and properties; equating test scores for using a common set of norms tables; approaches for addressing the goodness of fit between a latent trait model and a data set; (c) comparing the one-parameter and three-parameter logistic models for ability estimation and decision-making with several test lengths and ability levels; (d) determining the optimal length of criterion-referenced tests with different types of item pools (varying the level of item heterogeneity) and using two different item selection methods; (e) a system to allow instructors to specify required level of measurement precision and obtain information to help them determine test length; (f) comparing the fit of the one-parameter and three-parameter models to 25 sets of test data; and (g) building banks of valid test items. (Author)

Descriptors :   *Performance tests, *Teaching methods, *Mathematical models, Test methods, Instructors, Students, Adaptive training, Learning curves, Response, Skills, Performance(Human), Scoring, Decision making, Statistical decision theory, Computerized simulation, Air Force training

Subject Categories : Humanities and History
      Psychology
      Statistics and Probability

Distribution Statement : APPROVED FOR PUBLIC RELEASE