曝光率控制對多向度電腦化適性測驗能力估計信度之影響--以2001年國中基本學力測驗資料為例

Influences of Item Exposure Control on Reliability of Ability Estimation in Multidimensional Adaptive Testing: Using the Empirical Data of the 2001 Basic Competency Test for Junior High School

本研究探討曝光率控制方式對多向度電腦化適性測驗（Multidimensional Computerized Adaptive Testing， MCAT）能力估計信度的影響。具體而言，我們使用三種曝光率控制（嚴格控制、寬鬆控制、不控制）對2001年「國民中學學生基本學力測驗」之資料進行不同終止題數的題間多向度電腦化適性測驗（Between-item Multidimensional Computerized Adaptive Testing， BMCAT）。曝光率控制是以Sympson與Hetter（1985）法來進行。結果顯示：1.該控制法效果良好；2.當進行高終止題數的BMCAT時，曝光率控制對該測驗的能力估計信度沒有影響；而在低終止題數時，嚴格的曝光率控制會使信度略微降低0.04，顯示其影響並不大。因此在進行BMCAT時可以放心地加入曝光率控制以提高試題的安全性。

The influences of item exposure control on the reliability of ability estimation in Multidimensional Computerized Adaptive Testing (MCAT) are investigated by using the empirical data of the 2001 Basic Competency Test (BCTEST) for Junior High School. Three levels of item exposure rate control were manipulated under the long-test and short-test conditions in Between-item Multidimensional Computerized Adaptive Testing (BMCAT). Sympson and Hetter’s (1985) method was applied to control the item exposure rate. The Results show (a) the item exposure rates were successfully controlled， and (b) the reliability of ability estimation was not affected by the item exposure control under the long-test condition， but was reduced a little bit under the long-test condition. Implementa?tion issues in multidimensional adaptive testing are discussed.