| Summary: | 碩士 === 國立臺北大學 === 統計學系 === 94 === For most of the professional tests, language ability tests and other tests which offered periodically, it is important to equate the tests of the different periods. The existing designs for equating may not suitable for all the situation. In this study, we focus on the case that two tests in different seasons with some amount of the same examinees.
Data used in this study is provided by a famous professional institution. Two tests were offered four months apart. There are about one part of examinees are the same. The tests are multiple choices and there are dichotomous responses to each question. In this study, we discussed the one-parameter, two-parameter, and three-parameter logistic models in the item responses theory as the foundation of the learning effect correction. We, then, modified the existing IRT connecting designs based on the characteristic of this data and suggested a suitable equating type. Further, we discuss how to make up the examination under the same calibration. In this research, we set up a sufficient model of test analysis and offer some suggestions for the formulating questions of a test.
|
|---|
