NotesFAQContact Us
Search Tips
ERIC Number: ED476863
Record Type: Non-Journal
Publication Date: 2003-Apr
Pages: 49
Abstractor: N/A
Reference Count: N/A
Exploring Alternative Characteristic Curve Approaches to Linking Parameter Estimates from the Generalized Partial Credit Model.
Roberts, James S.; Bao, Han; Huang, Chun-Wei; Gagne, Phill
Characteristic curve approaches for linking parameters from the generalized partial credit model were examined for cases in which common (anchor) items are calibrated separately in two groups. Three of these approaches are simple extensions of the test characteristic curve (TCC), item characteristic curve (ICC), and operating characteristic curve (OCC) methods that have been previously developed for other binary item response models. The ICC approach explicitly provides a symmetric solution for estimating linking constants whereas the TCC and OCC approaches yield an asymmetric solution. Thus, the symmetry of the result is confounded with the type of characteristic curve used to derive the result. New characteristic curve techniques are developed to estimate this confound. Specifically, symmetric versions of the TCC and OCC methods are developed within the context of the generalized partial credit model (GPCM) along with an asymmetric version of the ICC technique. The accuracy of linking constant estimates and the accuracy of rescaled GPCM parameter estimates obtained with each method was examined in a simulation study. The study suggested that the TCC method yields slightly more accurate estimates of linking constants and model parameters as do symmetric, as opposed to asymmetric, solutions. The study suggests that all of the methods yield similar linking results when GPCM parameters are estimated accurately using large samples. (Contains 4 tables, 9 figures, and 13 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