本文探討在平行實驗設計下之過離散計數型資料的樣本數計算問題。我們使用經適當修正後，具強韌性的負二項概似函數，發展出一套計算樣本數的方式，此強韌有母數方法在不需知道資料真正分配的情形下亦能提供正確的樣本數。因此，以此強韌有母數方法分析平行實驗設計下之過離散的計數資料與計算所需樣本數是較佳的選擇。 此外我們也以Zhu & Lakkis (2013)的概念為基礎，推得另一套求取所需樣本數的計 算方式，並將此方法所得之結果與我們提出的方法算的樣本數做比較。 ;This thesis proposes a way of calculating sample size that is suitable for general count data and over-dispersed count data in parallel designs. This robust sample size calculation task is accomplished by employing the negative binomial as the working model with a proper adjustment. We use various data configurations in simulation studies to demonstrate the merit of our new approach for calculating sample size. Contrasts are also made with the sample size method proposed by Zhu & Lakkis (2014).
