某一種新藥想成為專利藥的合格替代藥品必須通過非劣性檢定。目 前文獻中針對右方設限存活資料，是根據建構信賴界限進行非劣性檢定。本文根據Freitag 等人(2006)提出的逐點檢定及最小上界檢定、Parzen 等人(1997)提出一個模擬方法及McKeaague 和Zhao(2002)以經驗概似函數建構之四個不同的信賴下界及在資料符合比例風險假設下，用Cox 模型推導出信賴下界。本文藉由模擬研究探討所提的五種信賴下界的覆蓋機率及與真實質之差距。最後，以一個實例說明上述五種信賴下界之應用。 A new drug to become a alternative medicine of a qualified patent medicine must be through non-inferiority test. Currently literature for right censored survival data is based on constructing a confidence bound for non-inferiority test. In this article, we proposed four different confidence bounds based on Freitag et al (2006) proposed pointwise test, and supremum test, Parzen et al (1997) proposed a simulation method, McKeaague and Zhao (2002) has constructed empirical likelihood function. And the data meet the proportional hazards assumption, using Cox models derived a confidence bound. A simulation study is then conducted to compare the coverage probability and the difference between the proposed confidence bounds and the true value. Finally, the application of the proposed confidence bounds is illustrated by using a real data.
