以SIBTEST改進隨機效果對檢測題組DIF之影響

A Modification of SIBTEST to Reduce the Influence of Random Effects in Detection of Testlet DIF

題組反應模式中的隨機效果會影響受試者的答對機率，而導致題組DIF無法被正確的檢測出來。本文即是以SIBTEST為對象，提出調整方法來減少隨機效果的干擾。此方法是在隨機效果保持不變與設定為零的兩種情況下，依照SIBTEST所得配對子測驗分數的差異進行迴歸分析，再以迴歸係數重新計算題組DIF估計值，接著應用bootstrap推導估計誤差及檢定統計量。模擬研究因素包括：隨機效果變異量、題組數、DIF試題數、受試者人數和能力差異等。結果顯示：調整方法不會受到隨機效果的影響，其型一錯誤率大都接近顯著水準0.05；而檢定力也都優於SIBTEST所得。此方法的確可以作為測驗分析者檢視題組DIF的參考。

Random effects usually affect the probability of a correct answer to testlet items. Therefore， the potential testlet DIF may not be identified appropriately. This study proposes a modification of SIBTEST to reduce the influence of random effects. The method applies the linear regression to calculate DIF estimates as well as the bootstrap to derive the test statistic. Factors used in the simulation study are the variance of random effects， the number of DIF items， the sample size， and the group ability difference. Results indicate that the refined method is able to detect testlet DIF with good powers while the type I error rates are approximated to the significant level of 0.05.