摘要 所謂的「因材施教」即是說明教師必須先瞭解每一位學生的長處與短處，才 能夠設計教學方針，實施補救教學。然而，一般傳統的紙筆測驗，僅提供學生在 團體中的量尺分數，並無法顯現出學生是否精熟某種概念的訊息。 為了進一步幫助學生或教師對於測驗的結果有更多的瞭解，進而施行更有效 率的學習，認知診斷模型可以提供解決的方案，而其中又以DINA 模式最簡單 也最容易解釋，目前國外已有許多學者投入模式的開發與實際應用的研究。同樣 地，參數不變性在認知診斷評量的測驗設計上著實具有相當重要地位，也就是可 以研究如何以認知診斷模型來估計試題參數的潛在基本特徵與其分佈變化。 本研究透過模擬樣本資料的實驗設計，探討在不同樣本人數、不同認知屬性 分佈之下，分別以DINA 模式與 G-DINA 模式估計來進行探討，當資料符合 DINA 模型時，有很明顯的參數不變性，若資料型態為未知時，建議採用G-DINA 模式來進行分析。

Abstract "Teaching students in accordance with their aptitude" means that teachers must understand each student's strengths and weaknesses so that they can design principles of teaching and remedial education. However, the general tradition written test is only supply scale score of each student in a group, but it cannot display if the students good at some kinds of skills. In order to help students and teachers to know about the test results for having more efficiency learning, the "cognitive diagnosis models" could provide the solutions. The DINA model is the simplest and easiest to explain among these models. Many foreign scholars are involved in the development and practical application research at the present time. Similarly, invariance of parameter is really important to the designing of cognitive diagnosis assessment, it means we can research how to use cognitive diagnosis models to estimate the potential basic features and the distributive changes of the item parameters. This paper adopt simulated data from design of experiment to explore that to estimate the parameters with DINA model and G-DINA model respectively so that we can compare with the existence of invariant in different sample size, and cognitive attribute distributions. The results show that the DINA and G-DINA model are invariant when the DINA model fits the data and the G-DINA model estimations get better accuracies when we don’t know about the data.