Nonparametric Hypothesis Testing: Rank and Permutation Methods with Applications in R

Presentation of the book

The importance and usefulness of nonparametric methods for testing statistical hypotheses has been growing in recent years mainly due to their flexibility, their efficiency and their ease of application to several different types of problems, including most important and frequently encountered multivariate cases. By also taking account that with respect to parametric counterparts they are much less demanding in terms of required assumptions, these peculiarities of nonparametric methods are making them quite popular and widely used even by non-statisticians.

The growing availability of adequate hardware and software tools for their practical application, and in particular of free access to software environments for statistical computing like R, represents one more reason for the great success of these methods.

The recognized simplicity and good power behavior of rank and permutation tests often make them preferable to the classical parametric procedures based on the assumption of normality or other distribution laws. In particular, permutation tests are generally asymptotically as powerful as their parametric counterparts in the conditions for the latter. Moreover, when data exchangeability with respect to samples is satisfied in the null hypothesis, permutation tests are always exact in the sense that their null distributions are known for any given dataset of any sample size. On the other hand, those of parametric counterparts are often known only ...