具差異試題功能之定錨試題對測驗等化之影響

Abstract

首先，依定錨試題中具差異試題功能比例不同而言，在測驗等化的過程中若定錨試題中的DIF試題所佔比例愈高則對受試者能力值的估計精準度誤差愈大。 其次在定錨試題中具差異試題功能難度值差異比較方面，受試者能力值經由等化後估計誤差會隨著試題難度值的增加而遞增，但其幅度並不明顯。 若比較受試者能力值差異對等化估計精準度之影響，研究中兩群受試者能力值差異設為0.5及1兩種，當在其他參數條件相同，其估計誤差會隨著能力值差異變大而使估計的誤差增加。 在單參數及三參數兩種不同等化模式之比較，若其他參數設定相同情況下，以三參數模式等化後所造成的能力值估計誤差明顯高於單參數估計所造成的誤差。 The purpose of this study is to investigate the effects of anchor items exhibiting DIF on test equating under the one parameter and three parameter logistic models. Four independent variables were manipulated: (a) item response models: the Rasch model and the 3-parameter logistic model: (b) percentage of DIF items in the anchor test: 10%, 20%, 30% and 40%; (c) magnitude of DIF: 0.4 and 0.6; (d) difference of mean ability between reference and focal groups: 0.5 and 1.0. The dependent variable was the MSE of latent trait estimation. Through the Monte Carlo studies, MSE were increasing when (1) the percentage of DIF items in the anchor test increasing: (2) the magnitude of DIF becomes larger; (3) the difference of mean ability between groups increasing: The latent trait parameters recovery were better under Rasch model than 3PL model.