Some Remarks On Statistical Inference

David R. Cox
Nuffield College, Oxford, UK


The variety of applications of statistics is so great that there is
some danger in trying to encompass them all within one framework.
Critical aspects in applications moreover commonly concern formulation
of the research question of interest rather than mode of inference per se.
Nevertheless it will be argued that for most purposes ideas centered on
assessing uncertainty via hypothetical conditional error rates, while
there are some problems in setting out a precise formulation, are appropriate.
Some comments on the role of Bayesian formulations will be added.