Model-free data condensation (Q3977945)

One cannot do better, in commenting on this paper, than quoting from the author's summary, viz. ``One aim of data analysis is its condensation, namely, capturing its gist in an apposite way. This paper addresses the problem of constructing and assessing such condensations without reference to mechanisms which might have generated the data. The results obtained lead to non-probabilistic interpretations of some well-known inferential procedures of classical statistics and thereby shed new light on the structure of statistical inference and the theory of probability.'' The range of topics discussed is wide, covering, among others, exponential families, Bayes' postulate, entropy and likelihood, computability, consistency and closure, the likelihood principle, and survival data. Almost every statistician or probabilist will find something of relevance in this paper, and its study will yield rich rewards.