在二元配對資料中，經常根據受測對象接受新舊兩種處理的成功機率之差異的信賴下界或上界，檢定新處理相對於舊處理是否具備非劣性或優越性。本文則根據此種成功機率比率進行上述檢定。應用Buehler方法修正邊際機率比率的三種近似信賴上界，使其最小覆蓋機率不低於信心水準，獲得此一比率的正確信賴上界。透過數值分析及模擬研究，在適當的評估標準之下，修正後的信賴域會優於修正前的信賴域。本文以數筆資料說明所提修正方法在非列性或優越性檢定的應用。 To that for the non-inferiority or superiority of a new treatment relative to the standard treatment based on binary matched pairs data, the lower or upper bound for the difference between the two probabilities of success is usually constructed. In this article, we modify these approximate upper confidence bounds for the ratio of the two probabilities according to the Buehler method. Therefore, we obtain these exact upper confidence bounds for the ratio that can be applied to testing for the non-inferiority and superiority. Moreover, the modified confidence bounds are all better than the original ones for holding their confidence levels. Finally, we demonstrat the application of the propose confidence bounds based on some numerical data sets.
