國中基本學力測驗之DIF的實徵分析：以91年度兩次測驗為例

An Empirical Analysis of DIF on Basic Competence Test: A Case Example of 2002 BCT

本研究主要目的為探討國中基本學力測驗不同考區間存在的DIF現況，以91年度兩次測驗之實徵資料為例，採M-H法進行試題的檢定，同時搭配ETS DIF分類系統，針對DIF試題進行分類，欲呈現一客觀及多角度的DIF檢定報告。研究結果如下：一、不同考區間DIF檢定情況，大致可獲得滿意的結果，其中以第二比較組（基隆與嘉義）DIF情況較嚴重；科目間，以社會科第一次測驗DIF整體現象最高。二、91年國中基測中仍存有少數偏高DIF比率的科目，且多數存在著一致性的有利或不利於某考區的現象。三、採M-H法進行DIF檢定時，同時搭配ETS DIF分類系統，較不受樣本因素影響，可得到一客觀、可信的DIF指標。最後根據研究結果，於文後提出相關建議與省思供讀者參考。

The purpose of this study is to investigate the phenomena of DIF on Basic Competence Test (BCT) among different exam-areas. We tried to use Mante-Haenszel method and ETS DIF classification rules to analyze the actual data of 2002 BCT two tests to present an objective and multidimensional DIF report. The conclusions are as follows: 1. The condition of DIF among different exam-areas may come to a satisfying result. Among exam-areas， the second comparison group (Keelung vs. Chiayi) exhibits the highest level of DIF; among exam-courses， the first social exam-course test exhibits the highest level of DIF. 2. There are still a few exam-courses having higher DIF ratio on 2002 BCT and mostly tending to favor or disfavor some exam-areas. 3. The use of M-H method along with ETS DIF classification rules may efficiently exclude the impact of sample sizes and obtain an objective-and-trustworthy DIF evidence. Finally， some suggestions and reflections for future studies are proposed.