## 期刊論文

 學年 96 2 2008-06-01 應用群序檢定方法於比較多個處理之試驗 Group Sequential Tests in Multi-Armed Trials 曾家誼; Tseng, Chia-yi; 林光男; Lin, Kuang-nan; 陳怡如; Chen, Yi-ju; 林國欽; Lin, Kuo-chin 淡江大學統計學系 臺灣智慧科技與應用統計學會 智慧科技與應用統計學報6(1)，頁29-38 針對臨床試驗的累積資料，常使用期中分析以有機會提早結束試驗，而群序檢定為常見的統計分析方法。本文將探討Pocock(1977), O'Brien-Fleming (1979）與三種不同顯著水準支配函數α1(t)=αt,α2(10,t)=α［(1-e(上標 -10t)/(1-e(上標 -10)］α3(-10,t)=α［(1-e(上標 10)］所產生的臨界值之差異，並討論其所需的樣本數與固定樣本數之比較，同時利用 Pocock 提出各階段名目顯著水準以α'的概念，計算出不同階段下的群序卡方檢定和群序F檢定之臨界值，此計算過程比 Jennison 與 Turnbull(1991)所提出的方法較為簡易並且其臨界值結果非常近似。此外，本文著重討論多個處理之群序檢定過程，以實例說明比較三種處理之群序檢定，同時採用Bonferroni 及 LSD 方法進行多重比較。 For accumulated data in clinical trials, three common group sequential methods were proposed by Pocock (1977), O’Brien and Fleming (1979), and Lan and DeMets (1983). The boundaries of Pocock and O'Brien-Fleming as well as three alpha spending functions:α1(t)=αt, α2(10,t)= α[(1-e^(-10t))/(1-e^(-10))], α3(-10,t)=α[(1-e^10t)/(1-e^10)] are discussed. We adopt the concept of nominal significance levelα'presented by Pocock to calculate the boundaries of group sequential chi-squared test and group sequential F test for a variety of overall significance levels and testing stages, which result in the similar boundaries of chi-squared test computed by Jennison and Turnbull (1991). The required sample size for each treatment, and average sample size for those methods are compared with the fixed sample size. Furthermore, the group sequential F procedure for multi-armed trials and the following multiple comparisons are illustrated by an example. 顯著水準支配函數; 臨界值; 多個處理之試驗; 多重比較;Alpha spending function; boundary; multi-armed trials; multiple comparison zh_TW 1812-433x 機構典藏連結 ( http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/20634 ) 機構典藏連結