No label defined (Q4805589)

Equivalence tests deal with problems where the alternative hypotheses specify only a small neighbourhood of the target parameter which indicates perfect coincidence of the distributions to be compared. This situation arises when one wishes to to verify a special parameter. In equivalence testing usually large samples are required to maintain some given bounds of both types of error risk. Therefore it is important to use tests satisfying rather strong optimality criteria. The theoretical basis for the construction of such tests has been laid already in the nineteen fifties. But up to now only one monograph exists addressing this field of problems. It is a German written text by the present author [Statistische Methoden zum Nachweis von Äquivalenz. (1994; Zbl 0798.62001)]. The first part of present book overlaps with that. About one third is new. After an introduction, the second chapter deals with `Methods for one-sided equivalence problems'. But the core of this monograph deals with methods of testing for equivalence in the strict, i.e., two-sided sense. `General approaches to the construction of tests for equivalence in the strict sense' are explained in chapter three. The next chapters are on equivalence tests for selected one-parameter problems, for designs with paired observations, and for two unrelated samples. `Multisample tests for equivalence' and `Tests for establishing goodness of fit' are chapters seven and eight. Especially the last chapter on `The assessment of bioequivalence' adresses a topic of overwhelming practical importance. The appendix gives a list of special computer programs, mainly in the SAS language, but some are Fortran programs. They are available at the WWW. The style is rather formal. Many examples illustrate the methods.