The Study of the Accuracy and Consistency of a Series of IRT Observed Score Equatings Using Various Scale Transformation Methods
Chien-Ming Cheng · 測驗學刊 · 2013
In order to evaluate equating accuracy and consistency, a chain of IRT observed score equatings were compared over different scale transformation approaches using simulated data. The study examined a series of equating results from separate and partial concurrent item parameter estimation, the results from the characteristic curve and the moment scale transformation methods, and the results from the chained equipercentile (CE) method. In addition, the study also compared the results from equating a new form directly to the base form (direct equating) to the results from equating the new form indirectly to the base form through some intermediated forms (indirect equating). The study showed: (i) IRT observed score equatings were consistent over different approaches to item parameter estimation and scale transformation; (ii) Direct equating functions were as accurate and consistent as indirect equating functions; (iii) The CE equating functions were as accurate and consistent as the IRT observed score equating functions at most score levels except the scores near the bottom and top levels.