漸進隨機選題之最大優先指數法在電腦化適性測驗之應用與選題控制成效之檢驗

Effect of the Progressive Maximum Priority Index Method on Item Selection Constraint in Computerized Adaptive Testing

本研究旨在發展一個新的電腦化適性測驗選題方法，稱之為「漸進隨機選題之最大優先指數法」（progressive maximum priority index method， PG-MPI），以期能同時達到選題條件限制的控制、題庫安全性的維持，與提升題庫使用率。透過一系列的模擬研究，操弄的情境包含題庫大小與測驗題長，並比較PG-MPI 與其他四種選題法（MPI法、PG法、最大訊息法與隨機選題法）之差異。結果發現在11 種限制條件之下，MI法與PG-MPI法能提供最佳的選題限制之控制，且能完全控制最大試題曝光率於設定的標準內；當加速度參數設定較低的數值時，PG-MPI 能提供尚佳的能力估計精準度；PG 法與PG-MPI 法能提高題庫的使用率。綜合上述發現，PG-MPI 法兼顧PG 法與MPI 法的優點，不但能滿足所有選題限制，維持題庫安全、提升題庫使用率，更能保有不錯的能力估計精準度。最後，研究者也提出若干建議供未來研究參考。

This study aimed to develop a procedure for severely constrained item selection with improving item pool usage rate in computerized adaptive testing. The authors proposed the progressive maximum priority index method (PG-MPI) to combine the maximum priority index method (MPI) with the progressive method (PG). Through a series of simulation， independent of pool size and test length， it showed that compared with the other four item selection methods the PG-MPI is able to accommodate various nonstatistical constraints， lead to fewer constraint violations， provide better pool security maintaining， and increase the item bank utility simultaneously with little loss in measurement precision when the acceleration parameter is set to lower value. Finally， study limitations are noted and suggestions for future investigations are proposed.