測驗或是評比皆會發生同分的現象，此時通常會利用某些主觀的程序來決定優先錄取的順序。本研究試圖提出一套較為客觀的程序，解決這個難題。本研究基本的理念是：考生的作答反應愈不合理，就表示有愈多跟測驗無關的因素在干擾其作答，其作答反應就愈不可靠，因此其錄取的順序就應愈後面。在 Rasch 分析裡，透過 fit 統計量可以判斷考生作答反應的合理性。我們利用 Quest 電腦軟體，分析500考生回答37道選擇題的測驗資料，從中抽取2組各3位得分相同的考生，操弄其異常的作答反應，然後評估各項fit統計量的敏感度。結果發現，outfit MNSQ的敏感度較佳，其值愈小表示反應愈合理。當 outfit MNSQ 相同時，可再考量以 infit MNSQ 為第二階段的辨別依據。除此之外，本研究還製作了網路版的 KIDMAP，可圖解分析考生的逐題反應情形，這將使得學習評量除了用以比較個別差異外，還可以發揮個別化的診斷功能。

It is very common that examinees have identical test scores. In such a case, some subjective procedures are often used to determine admission priority for those exami-nees with identical scores. This study proposed an objective procedure to determine the priority. The basic rational was: the more the unexpected item responses, the more the extraneous resources in the item responses, the less reliable the score, and thus the low-er the priority should be. In Rasch analysis, the fit statistics are used to assess the ex-pectedness of item responses and thus can be used to determine the priority. A real data set of 37 multiple-choice items and 500 examinees was analyzed using Quest. Two gro-ups, each with 3 examinees having identical total scores, were selected and their item responses were manipulated to be unexpected. The results showed that the outfitMNSQ
was more sensible to extremely unexpected responses, and the smaller the outfit MNSQ, the more reliable the item responses and thus the higher the priority should be. It was also suggested that if the outfit MNSQ statistics are identical, the infit MNSQ statistics can then be used to determine the priority. In addition, a web-KIDMAP was developed to provide easy inspection of the priority and to facilitate visual analysis of
individual item responses, such that assessment can become more individualized and diagnostic.