Multiple testing

Abstract

Abstract Multiplicity of statistical inference often arises in applications. This chapter presents and discusses methods for dealing with multiple testing. Alternatives to the simple but conservative Bonferroni correction are especially useful when dealing with small data sets when very small p-values are hard to obtain. The alternative methods introduced in this chapter include the Bonferroni-Holm method, hierarchical hypothesis testing, gatekeeping procedures, and closed testing procedures. Furthermore, a composite endpoint is discussed as an alternative to a multiple testing procedure. The family-wise error rate and the false discovery rate are considered. Simultaneous confidence intervals are also presented. R code is given.