Westfall Peter H
Department of ISQS, Texas Tech University, Lubbock, Texas 79409-2101, USA.
J Biopharm Stat. 2011 Nov;21(6):1187-205. doi: 10.1080/10543406.2011.607751.
There are many ways to bootstrap data for multiple comparisons procedures. Methods described here include (i) bootstrap (parametric and nonparametric) as a generalization of classical normal-based MaxT methods, (ii) bootstrap as an approximation to exact permutation methods, (iii) bootstrap as a generator of realistic null data sets, and (iv) bootstrap as a generator of realistic non-null data sets. Resampling of MinP versus MaxT is discussed, and the use of the bootstrap for closed testing is also presented. Applications to biopharmaceutical statistics are given.
对于多重比较程序，有多种用于自抽样数据的方法。这里描述的方法包括：(i) 作为基于经典正态的MaxT方法推广的自抽样（参数化和非参数化）；(ii) 作为精确置换方法近似的自抽样；(iii) 作为现实零假设数据集生成器的自抽样；以及(iv) 作为现实非零假设数据集生成器的自抽样。讨论了MinP与MaxT的重抽样，并介绍了自抽样在封闭检验中的应用。还给出了在生物制药统计学中的应用。
