nparcomp: An R Software Package for Nonparametric Multiple Comparisons and Simultaneous Confidence Intervals
nparcomp: An R Software Package for Nonparametric Multiple Comparisons and Simultaneous Confidence Intervals
复制标题
DOI：
10.18637/jss.v064.i09
复制
发表时间：
2015-03
影响因子：
5.8
通讯作者：
F. Konietschke;Marius Placzek;F. Schaarschmidt;L. Hothorn
中科院分区：
文献类型：
--
作者：
F. Konietschke;Marius Placzek;F. Schaarschmidt;L. Hothorn
单向布局，即单个因素具有多个层次，并且每个层次上有多个观察值，在各个领域中经常出现。通常，不仅对全局假设感兴趣，而且对不同治疗水平之间的多重比较也感兴趣。在大多数实际情况下，观测数据的分布是未知的，并且可能存在许多非典型测量值和异常值。因此，使用对观察样本施加限制性分布假设的参数和半参数程序变得值得怀疑。这反过来又强调了对统计程序的需求，使我们能够在可用数据的最小条件下准确可靠地分析单向布局。非参数方法提供了这种可能性，因此变得特别重要。在本文中，我们介绍了一种新的 R 包 nparcomp，它提供了一种简单且用户友好的方式来访问基于排名的方法，以分析不平衡的单向布局。它提供了执行多重比较并计算同时置信区间的程序，以轻松可视化估计效果。包括两个样本的特殊情况，即非参数 Behrens-Fisher 问题。我们通过生物学和医学的例子来说明所实施的程序。
One-way layouts, i.e., a single factor with several levels and multiple observations at each level, frequently arise in various fields. Usually not only a global hypothesis is of interest but also multiple comparisons between the different treatment levels. In most practical situations, the distribution of observed data is unknown and there may exist a number of atypical measurements and outliers. Hence, use of parametric and semiparametric procedures that impose restrictive distributional assumptions on observed samples becomes questionable. This, in turn, emphasizes the demand on statistical procedures that enable us to accurately and reliably analyze one-way layouts with minimal conditions on available data. Nonparametric methods offer such a possibility and thus become of particular practical importance. In this article, we introduce a new R package nparcomp which provides an easy and user-friendly access to rank-based methods for the analysis of unbalanced one-way layouts. It provides procedures performing multiple comparisons and computing simultaneous confidence intervals for the estimated effects which can be easily visualized. The special case of two samples, the nonparametric Behrens-Fisher problem, is included. We illustrate the implemented procedures by examples from biology and medicine.