ERIC Number: ED457187
Record Type: Non-Journal
Publication Date: 2001-Apr
Equating Parameter Estimates from the Generalized Graded Unfolding Model.
Roberts, James S.
Three common methods for equating parameter estimates from binary item response theory models are extended to the generalized grading unfolding model (GGUM). The GGUM is an item response model in which single-peaked, nonmonotonic expected value functions are implemented for polytomous responses. GGUM parameter estimates are equated using extended versions of the mean-sigma, mean-mean, and item characteristic curve methods. The former two methods are implemented using two different strategies based on alternative parameterizations of the GGUM. All of these methods attempt to estimate a scale constant (A) and a location constant (B) that can equate the metric of item response model parameters derived from separate calibrations. A small simulation is performed to provide preliminary information about the characteristics of the alternative equating methods studied. The item characteristic curve method performed best with regard to the mean squared error, bias, and standard error of equating constant estimates as well as the absence of extremely deviant estimates. It was noted that, although the average superiority of estimates produced by the item characteristic curve method was quite small, substantial outliers sometimes emerged when estimating equating constants with other methods. Consequently, the item characteristic curve method is recommended as a means to develop estimates of equating constants in the GGUM. An appendix discusses equating parameter estimates from the GGUM. (Contains 3 figures, 1 table, and 17 references.) (Author/SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A