The Impact of Parameter Estimation Error on IRT Linking Methods with Polytomous Items
The information-weighted characteristic curve methods have shown excellent performance in IRT linking with dichotomous i-tems.However,few researches explore the effect of parameter estimation error on test linking.This paper extends the information-weighted characteristic curve methods to IRT linking with polytomous items and explores the effects of parameter estimation error,ability differences,and test length on linking through simulation studies.IRT linking performance was evaluated using indices related to charac-teristic curves and errors.The results indicated that the information-weighted characteristic curve methods performed slightly better than the traditional characteristic curve methods,while other new methods performed as well as the traditional methods.The linking per-formance was better when the parameter estimation error and ability differences were smaller,and the test was longer.The bias and vari-ance tradeoff provides a new direction for test linking and equating.
万方数据
维普
