| Summary: | 碩士 === 國立臺中教育大學 === 特殊教育學系碩士在職專班 === 102 === This study aims at developing a suitable computerized adaptive test of reading comprehension for fifth graders and saving administration items and time. Subjects were 491 fifth graders recruited from six elementary schools of central regions. The obtained data were analyzed by the classical test theory (CTT), the item response theory (IRT) and the examination of reliability and validity. The results are as followed:
1. The Cronbach’s α of the test was .899, showing good internal consistency.
2. When the data were analyzed by the classical test theory, the outcome showed that the difficulty of the present test was within the range of medium and easy featuring ideal discrimination index.
3. When the data were analyzed by the item response theory, the outcome showed that the difficulty of the present test was within the range of medium and easy featuring acceptable conjectural variation and ideal discrimination index.
4. When the data were analyzed by the EAP and MAP methods simulating operations of the computerized adaptive test, the hypotheses of saving test items and administration time are accepted.
5. Statistical methods of maximum likelihood estimation (MLE), maximum a posteriori (MAP) and expected a posteriori (EAP) were verified simulating operations of the computerized adaptive test. The results showed that maximum likelihood estimation (MLE) and expected a posteriori (EAP) achieved better results.
|
|---|
