逐步檢定程序可用來判斷新治療方法是否比k個原有的治療方法優越,或具有相等性效力,由於一般在使用逐步檢定時,往往假設常態母體,因此本文主要探討母體不為常態分配的情況下,如何將無母數檢定統計量應用於逐步檢定程序。由數值範例得知在常態母體下,判斷新治療方法是否與k個原有的治療方法具有優越性或相等性時,使用無母數逐步檢定方法與原有T逐步檢定方法的結論相同。此外,利用模擬結果顯示在非常態母體下,無母數逐步檢定方法的型Ⅰ族群誤差率可控制在所設定的顯著水準範圍內,且其檢定力比T逐步檢定方法的檢定力較高。
Stepwise testing procedures can be used to establish superiority and equivalence of a new treatment with respect to each of the k standard treatments. The regular assumptions of the stepwise testing procedures are normality. The objective of the present paper is to release the assumption and replace the t test statistic by nonparametric test statistics in stepwise testing procedures. The numerical example based on normality shows that the proposed nonparametric testing procedures and the multiple testing procedures using T test statistic reach the same conclusions. Simulation studies indicate that the proposed nonparametric testing procedures coincide the required condition that type Ⅰ familywise error rate is less than the specified significance level, and the proposed procedures have higher power than the stepwise T testing procedures under nonnormality.
關聯:
智慧科技與應用統計學報=Journal of Taiwan Intelligent Technologies and Applied Statistics 4(2),頁 115-134