具作答回溯機制之認知診斷電腦化適性測驗   全文下載

Reviewable Cognitive Diagnostic Computerized Adaptive Test

許多研究顯示，具作答回溯機制之電腦化適性測驗比不具作答回溯機制之電腦化適性測驗在能力估計上更精確。目前具作答回溯機制之電腦化適性測驗於試題反應理論上開發，並設計提供受試者潛在連續特徵的總結報告，而非提供足夠的診斷訊息。認知診斷模型可估計受試者認知屬性的精熟狀態，而此對補救教學具有顯著的助益。本研究基於Deterministic Inputs, Noisy“And”Gate model（DINA模型）來開發具動態區塊回溯機制之認知診斷電腦化適性測驗。透過模擬Q矩陣、試題參數、試題回溯模式、回溯區塊數量、回溯行為、回溯機率等各項設定來探討本研究之效益。在大部分的情形下，研究顯示可比傳統不具作答回溯機制之認知診斷電腦化適性測驗提供更高的估計正確率。

Many studies showed that the ability estimate from the reviewable computerized adaptive test would be more précised than that from the non-reviewable computerized adaptive test. Current reviewable computerized adaptive tests were developed based on item response theory and were designed to offer the examinee a summary report of a latent continuous trait rather than to provide sufficient diagnostic information. Cognitive diagnostic models could estimate the mastery status of the examinees’ cognitive attributes which would have a great benefit to the remedial instruction. The purpose of this study was to develop a dynamic block-review computerized adaptive testing algorithms based on Deterministic Inputs, Noisy “And” Gate model (DINA model). The performance of the proposed algorithms was investigated by a simulation study with various settings, Q-matrix, item parameters, item review mode, size of review block, review behavior, and review probability. Under most situation, the results of this study indicated that the proposed algorithm provided higher estimation accuracy than the traditional non-reviewable cognitive diagnostic computerized adaptive test.