NotesFAQContact Us
Search Tips
ERIC Number: ED454267
Record Type: Non-Journal
Publication Date: 2001-Apr-13
Pages: 37
Abstractor: N/A
Reference Count: N/A
Reliability of Scores from Tests Composed of Testlets: A Comparison of Methods.
Hendrickson, Amy B.
The purpose of the study was to compare reliability estimates for a test composed of stimulus-dependent testlets as derived from item scores, testlet scores, and under the univariate generalizability theory and multivariate generalizability theory designs, as well as to determine the influence of the number of testlets and the number of items per testlet on the generalizability coefficient. For the study, random samples of 3000 examinees were drawn from the standardization data of a large standardized test. As expected, item score reliability values were largest, while reliability based on testlet scores was lowest. Generalizability coefficient estimates from the univariate and multivariate designs fell between the item and testlet reliability estimates, yet were considerably smaller (about 0.03) than the item score estimates. The multivariate analysis incorporates all items and stimulus information to obtain the most accurate reliability estimate. Four appendixes contain MGENOVA code for some item results. (Contains 10 tables and 13 references.) (Author/SLD)
Publication Type: Numerical/Quantitative Data; Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A