Florida School of Professional Psychology, Argosy University, Tampa, FL, USA.
NeuroRehabilitation. 2013;32(3):693-6. doi: 10.3233/NRE-130893.
The traditional Bonferroni method is a commonly used post hoc hypothesis test to adjust for familywise error rate inflation; however, a less well-known derivative of this test, the Holm's sequential procedure, provides an alternative method for familywise error rate correction. This less conservative approach is particularly relevant to studies investigating neuropsychological functioning because of the extent to which neuropsychological datasets tend to include interrelated outcome measures, reducing the relevance of the universal null hypothesis assumption, on which the traditional Bonferroni method relies.
Mathematical illustrations demonstrating how to compute the two adjustments are provided. The two methods are compared using a simple hypothetical dataset.
By using the divisors (n - j + 1) in lieu of n, Holm's sequential procedure is guaranteed to never reject fewer hypotheses than the traditional Bonferroni adjustment.
The Holm's sequential procedure corrects for Type I error as effectively as the traditional Bonferroni method while retaining more statistical power. Although the Holm's sequential procedure is more complicated to compute than the traditional Bonferroni method, the Holm's sequential procedure may be a more appropriate method for adjusting familywise error rate inflation in many types of neuropsychological research.
传统的 Bonferroni 方法是一种常用的事后假设检验方法，用于调整总体错误率膨胀；然而，这种检验方法的一个不太知名的衍生方法，即 Holm 的序贯程序，提供了一种用于总体错误率校正的替代方法。由于神经心理学数据集往往包含相互关联的结果测量，因此这种方法不太保守，对于研究神经心理学功能尤其相关，这降低了传统 Bonferroni 方法所依赖的普遍零假设假设的相关性。
提供了数学说明，演示如何计算这两种调整。使用一个简单的假设数据集比较了这两种方法。
通过使用除数 (n - j + 1) 代替 n， Holm 的序贯程序保证不会比传统的 Bonferroni 调整拒绝更少的假设。
Holm 的序贯程序在保留更多统计功效的同时，有效地纠正了 Type I 错误。尽管 Holm 的序贯程序比传统的 Bonferroni 方法更复杂，但在许多类型的神经心理学研究中，Holm 的序贯程序可能是调整总体错误率膨胀的更合适方法。
