| Summary: | 碩士 === 國立臺灣大學 === 機械工程學研究所 === 84 === Two separate test forms are administered to two groups of examinees, such that each group is assigned one test form but not both due to practical constrains. Test equating between the two test scores distribution is a well-known incomplete data problem in educational statistics. The frequency estimation method, combined with the log-linear smoothing model, and their extension to EM algorithm data imputation have received much attention. All these methods make use of a reference common-item test scores, which are observed for both groups. In practice, it is often infeasible to administer a separate reference test, and the common-item scores may be selected form the examinees' past records, such as the junior or senior high school grades. On the other hand, groups may be randomly selected from certain strata of the entire population of interest.
|
|---|
