NotesFAQContact Us
Search Tips
ERIC Number: ED466639
Record Type: Non-Journal
Publication Date: 2002-Apr
Pages: 43
Abstractor: N/A
Reference Count: N/A
Comparing Vertical Scales Derived from Dichotomous and Polytomous IRT Models for a Test Composed of Testlets.
Bishop, N. Scott; Omar, Md Hafidz
Previous research has shown that testlet structures often violate important assumptions of dichotomous item response theory (D-IRT) models, applied to item-level scores, that can in turn affect the results of many measurement applications. In this situation, polytomous IRT (P-IRT) models, applied to testlet-level scores, have been used as an alternative. The objective of this study was to examine the distributional characteristics of vertical scales created using selected D-IRT and P-IRT models for a test composed of testlets. Iowa Tests of Basic Skills (ITBS) Reading Comprehension test scores from 60,000 randomly selected students (10,000 students per grade for grades 3 through 8) who took the ITBS during a recent fall administration were used in this study. For this data, vertical scales were produced using three D-IRT models (the Rasch model and the three-parameter logistic model calibrated under concurrent and separate group options) and four P-IRT models (the nominal model, the graded response model, the partial credit model, and the generalized partial credit model). The results show a number of differences in the distributional properties of vertical scales derived from IRT models applied at the item and testlet levels. (Contains 4 figures, 8 tables, and 53 references.) (Author/SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A
Identifiers - Assessments and Surveys: Iowa Tests of Basic Skills