本論文指出二元負二項模型可經適當的修正後提供具強韌性的概似函數。在不知道資料的聯合分配情況之下,可利用此強韌概似函數分析在交叉實驗設計下的個數資料,進而決定樣本數。我們並將此有母數的強韌概似函數法與Lui (2013) 所提的方法做有效性的比較。;The aim of this dissertation is to show that the bivariate negative binomial distribution, when properly adjusted, is a convenient tool for analyzing count data from cross over designs. We contrast this universal parametric robust with a currently proposed method in literature in terms of validity and efficiency. The issue of sample size determination is also examined.