ABSTRACT
In this final chapter we present a range of topics that go beyond our main focus
of likelihood-based inference for parametric models with large samples. The inten-
tion is not to provide a complete presentation of these topics but rather to give a
brief introduction to approaches that can be used without making parametric or large
sample assumptions. We begin with exact methods that make use of the actual sam-
pling distribution in small samples rather than normal or χ2 approximations for large samples. We then describe non-parametric methods that require no assumptions, or
at least very few assumptions, about the probability distribution that generated our
sample. We also discuss an important class of computer-intensive methods called re-
sampling methods, particularly bootstrapping. While the presentation is necessarily
cursory, our discussion provides a taste of the breadth of additional inference topics
required by practising biostatisticians. The methods are illustrated by reconsidering
a number of our earlier examples from these alternative perspectives.
