臺灣博碩士論文加值系統

　　由於Internet技術成熟，使得適性測驗得以網路化，受試者可不受時空限制，自遠端上線接受測驗。本研究旨在結合適性測驗（adaptive testing）及項目反應理論（item response theory, IRT），以三層式主從架構（three - tier client server architecture）在Internet上發展遠距適性測驗系統（distance computerized adaptive testing; DCAT）。
　　一般常用的成就測驗多採傳統的紙筆式（paper - and - pencil）施測，但對受試者而言，通常只有部份題目難度符合其能力，其餘題目皆太難或太容易。採用項目反應理論設計之電腦化適性測驗是依受試者程度，選取最符合受試者能力的試題施測，受試者每做完一題便重新評估能力值，再依適性策略（adaptive strategy）決定次一題目，如此受試者只需以較少的題數，便可以精確的估計出其能力值，此可有效的節省測驗成本，提昇效益。
　　本研究分為三個步驟：首先規畫一個具平臺獨立之遠距測驗系統架構；其次，設計題庫管理子系統、測驗管理子系統以及依據項目反應理論設計線上適性測驗子系統；最後實作系統於Internet的環境中。在client端因考慮到使用方便性、測驗安全性以及適性測驗需用到大量的數值計算，故採用Java設計以供不同平臺之系統使用。Server端則建置題庫、WWW server、以及DCAT server，負責處理client端要求，從題庫中擷取適當的題目施測，並將施測結果予以記錄。
　　經測試結果顯示本系統確可在Internet上，提供一穩定且有效率的遠距適性測驗環境，在結束條件計標準誤設定為0.4，0.3，0.2時，受試者平均施測題數為8.14，13.06，28.92題，模擬能力與估計能力之間的相關係數分別是0.9797，0.9877，0.9945，且經過pair t - test　結果顯示模擬能力與估計能力之間並無顯著差異，的確能有效縮短測驗長度又不失準確性。

　　Due to the maturing of Internet technology, the adaptive testing can be utilized the network and the examinee can take the test anywhere and any time. The purpose of the research is applying item response theory (IRT), adaptive testing theory and network technique to construct a system of distance computerized adaptive testing (DCAT) by the framework of three - tiered client server distance testing.
　　To operate the DC AT on the Internet has become a popular trend in the application in testing. Therefore, a system of DCAT in the World Wide Web environment is initiated in this study. This study consists of three stages. First, a framework of distance testing that has the characteristic of platform independence is constructed. Next, three subsystems are established. One is the on - line testing subsystem based on IRT on the client side, The second one is the test management subsystem. and the third one is database subsystem on the server side. Finally, a system of DCAT will be implemented into the Internet. To put the convenient usage of the examinee, the security of testing and the heavy numerical calculation into consideration, the Java is used to design a cross - platform program at client side. Besides, the server side must process the request from the client side and then select the appropriate item to test examinee and record the testing result.
　　When the stopping condition of the estimation of the examinee''s ability is assigned as 0.4, 0.3 or 0.2, the mean number of testing items are 8.14, 13.06, or 28.92. The correlation coefficients between the simulated and estimated ability under the three stopping conditions are 0.9797, 0.9877, or 0.9945, respectively. The results of test show that the system developed in this study can satisfy the need of reducing the number of testing items needed and still maintain the accuracy of the testing. Also the system developed here can provide a stable and efficient testing environment for distance adaptive test on the Internet network.