II Objectivity and Conditionality in Frequentist Inference

Statistical methods are used to some extent in virtually all areas of science, technology, public affairs, and private enterprise. The variety of applications makes any single unifying discussion difficult if not impossible. We concentrate on the role of statistics in research in the natural and social sciences and the associated technologies. Our aim is to give a relatively nontechnical discussion of some of the conceptual issues involved and to bring out some connections with general epistemological problems of statistical inference in science. In the first part of this chapter (7(I)), we considered how frequentist statistics may serve as an account of inductive inference, but because this depends on being able to apply its methods to appropriately circumscribed contexts, we need to address some of the problems in obtaining the methods with the properties we wish them to have. Given the variety of judgments and background information this requires, it may be questioned whether any account of inductive learning can succeed in being “objective.” However, statistical methods do, we think, promote the aim of achieving enhanced understanding of the real world, in some broad sense, and in this some notion of objectivity is crucial. We begin by briefly discussing this concept as it arises in statistical inference in science.