實驗設計目前已經被廣泛應用於各種領域當中,當安排實驗進行時會遇到因為時間、地點或其他原因,造成條理性的誤差,在這種情況下我們會使用集區設計,但在集區內的觀測值往往彼此之間並不互為獨立,因此在考量到集區內觀測值彼此相關的情況下,如何得出一個良好的實驗便成為目前所關注的問題之一。 本研究考量在集區內觀測值彼此之間互為相關之情形下如何利用一個有系統的方法來建立一個使主效應估計值的變異數皆相等的完全因子實驗與二分之一的部分因子實驗,並且提供範例及簡單的驗證。而使往後的實驗設計者可以只憑簡單的實驗設計知識便可以輕鬆的設計出一個良好的實驗。 In many situations, observations in factorial design experiments may be dependent. In this article, we provide a explicit method to construct balanced designs for 2n and 2n-1 factorial designs when observations in blocks are correlated with positive correlation. We first characterized the balanced design, and then illustrate the method through balanced 23 full factorial designs and 24-1 half replicated factorial designs , finally we explain why the design made by the method is a balanced design and provide an analytical proof of balanced designs for both 2n full factorial designs and 2n-1 half replicated factorial designs .