應用Mantel-Haenszel法於連續反應試題之差異試題功能檢核

Applying the Mantel-Haenszel Method to Exploratory Differential Item Functioning Assessment in Continuous Response Items

差異試題功能（differential item functioning, DIF）檢核對於測驗效度與測驗公平性的維護相當重要。在過去數十年，陸續有許多DIF檢核方法被提出，例如Mantel-Haenszel（MH）法等。然而，這些方法皆被應用於間斷反應試題（discrete response items）；對於連續反應試題（continuous response items）的DIF檢核，則尚未有研究提及。考量MH法在實務上的穩健表現，本研究進一步將Rayner與Best（2012）所推導出的連續反應MH統計量（稱之為MHC法），應用於連續反應試題的DIF檢核，並搭配量尺淨化（scale purification, SP）程序以提升DIF檢核效能。透過模擬研究顯示，MHC法搭配SP能有效控制型一誤差，同時具有良好的檢定力。由於此方法在執行上的難度並不高，本研究建議實務測驗應用者利用MHC法搭配SP對連續反應試題進行DIF檢核，以維護測驗品質。

Differential item functioning (DIF) assessment is critical for ensuring test validity and fairness. Many DIF assessment methods have been proposed in the past several decades, including the Mantel-Haenszel (MH) method. Such methods are applied to DIF assessment for discrete response items; however, no research has been done for continuous response items. Considering the popularity of the MH method in practical applications, its continuous counterpart (called the MHC method) proposed by Rayner and Best (2012) is applied to assess DIF for continuous response items in this study. The scale purification (SP) is further incorporated to improve the performance of the MHC method in DIF assessment. According to the simulation results, the MHC method with the SP procedure can yield high power rates while controlling type I error rates well in DIF assessment. Since the MHC method can be easily implemented with the SP procedure, it is recommended for test practitioners to conduct DIF assessments to improve the test quality for continuous response items.