| Summary: | 碩士 === 國立臺中教育大學 === 教育測驗統計研究所 === 98 === The study is based on Bayesian Networks and knowledge structure. The designs of lesson, how to formularize, as the learning content, compiling necessary digital individual instructional learning material and diagnostic test. By experimental design, it divides the teachers and students into 3 groups, experiment A （ a teacher paired with a student）, experiment B （ a teacher paired with 2 students） and Comparison （ a teacher paired with a group of students）. Based upon the procedures, we can explore the learning results with 3 respective instructive sets and digital individual instructive materials. It as well studies the question-saving ratio and accuracy of computerized adaptive diagnostic tests and the error pattern under adaptive sampling and anticipation consistency of sub-skills.
The facts of the study are listed as follows:
1. The diagnostic testing Cronbacha reliability on the lesson, How to formularize, goes to .87. The item difficulties are between .23 and .96. The item discriminations are all more than .23, most of them falling between .3 and .5. It is an excellent educational testing.
2. The computerized adaptive diagnostic tests can effectively save up to 42% of the questions and the accuracy goes to 96% or more.
3. Regarding the results of predictable consistence in the error pattern and sub-skill under the adaptive sampling, the predictable consistence of pretests is 93.36%; the one of posttests falls on 94.61%. The computerized adaptive diagnostic testing has high stabilities with the Bayesian network as the inferring mode.
4. On the learning effectiveness: The individual instructive mode and the group mode have the significant effect on pretests. The different score is 7.58** （**p<.01）; the individual instructive mode and the group mode have the significant effect on posttests. The different score is 3.46** （**p<.05）. The result reveals that digital individual instruction has better outcomes than traditional group learning.
|
|---|
