John W. Tukey's contributions to multiple comparisons (Q1873608)

In 1953 John W. Tukey (1915-2000) accomplished a mimeograph paper ``The problem of multiple comparisons'', which did much to shape the philosophy, mathematical development and practical applications of multiple inference. No other work or writer has had comparable influence. The proper treatment of multiplicity, which should take into account the ``trade-off between extracting belief from data and payment of error'' (Tukey 1991) is regarded by many as a critical component in a disciplined program of scientific research. Confidence intervals are identified with indication and significance tests with sanctification and action. He resumed intensive activity in this field during ihe last decade of his life. He introduced the notions of \textit{higher criticism} and of \textit{Wholly significant difference} and he was concerned with two forces: efficiency and robustness of Tukey's conjecture on unbalanced pairwise comparison problem. L. D. Brown (1979) proved the conjecture for the cases 3, 4, and 5. A. J. Hayter (1984) was able to prove the 31-years-old conjecture for all \(k\). Tukey's long running debate with David B. Duncan (1916) was more philosophical. His disagreement with Ronald A. Fisher (1890-1962) on the use of least significant difference was more methodological and concerned with the extraction of maximum useful information from the data at hand for a given simultaneous error-rate. His continuing argument with Henry Scheffé (1907-1977) on the use of \(F\)-projections had both philosophical and methodological components. Tukey was convinced that Statistics is a part of science and not a branch of mathematics only. 54 references.