Good's work in probability, statistics and the philosophy of science (Q922515)

In the collected papers of \textit{I. J. Good} [Good thinking: The foundations of probability and its applications (1983; Zbl 0583.60001)] an annotated bibliography with at least 1600 items to his credit and an 45-page index is given. Good had interrupted his studies during the second world war, he worked on cryptoanalysis under the famous Allan Turing (1912-1954) and his highly intelligent staff. The dominant feature present in almost all of Good's writing is his concept of probability, its meaning and its using. The ratio of posterior odds to prior odds which Good calls the Bayesian factor is equal to the ratio of the probability of the data on hypothesis H and on its negation \(\bar H.\) The Bayesian/non-Bayesian compromise is considered and the closely related principle of optimal stopping is frequently mentioned. Another example of Good's liberal thinking is his classification of Bayesians into 46656 varieties, this is based on 11 facets, for example facet 10 describes the Bayesian attitude to \(\sigma\)-additivity: a) finite additivity (as de Finetti), b) \(\sigma\)-additivity (almost all probabilists), c) using \(\sigma\)-additivity when mathematically convenient by regarding it as as barely relevant to the philosophy of applied statistics (Good and Lindley). Other topics are: weight of evidence, multinomials and contingency tables, hierarchical models, species estimation, density smoothing. For him all of life is interesting and amenable to clear thinking, as paranormal and natural selection, or reviewing a scholarly work on astrology.