Publication data: Princeton Mathematical Series, vol. 9. Princeton University Press, Princeton, N. J., 1946. xvi+575 pp. (A first version was published by Almqvist & Wiksell in Uppsala, Sweden, but had little circulation because of World War II.)
Description: Carefully written and extensive account of measure-theoreticprobability for statisticians, along with careful mathematical treatment of classical statistics.
Importance: Made measure-theoretic probability the standard language for advanced statistics in the English-speaking world, following its earlier adoption in France and the USSR.
Importance: Made Wald's ideas accessible. Collected and organized many results of statistical theory that were scattered throughout journal articles, civilizing statistics.
Description: In this paper Bayes addresses the problem of using a sequence of identical "trials" to determine the per-trial probability of "success" – the so-called inverse probability problem. It later inspired the theorem that bears his name (Bayes' theorem). See also Pierre Simon de Laplace.
Publication data: * Ramsey, Frank Plumpton; "Truth and Probability" (PDF), Chapter VII in The Foundations of Mathematics and other Logical Essays (1931).
Description: The first complete analysis of Bayesian Inference for many statistical problems.
Importance: Includes a large body of research on Bayesian analysis for outlier problems, variance components, linear models and multivariate statistics.
Publication data: Two volumes, A.F.M. Smith and A. Machi (trs.), New York: John Wiley & Sons, Inc., 1974, 1975.
Description: The first detailed statement of the operational subjective position, dating from the author's research in the 1920s and 30s.
Importance: Emphasizes exchangeable random variables which are often mixtures of independent random variables. Argues for finitely additive probability measures that need not be countably additive. Emphasizes expectations rather than probability measures.
Publication data: preliminary edition, 1965. Cambridge, Mass.: MIT Press, 1995.
Description: Extensive exposition of statistical decision theory, statistics, and decision analysis from a Bayesian standpoint. Many examples and problems come from business and economics.
Importance: Greatly extended the scope of applied Bayesian statistics by using conjugate priors for exponential families. Extensive treatment of sequential decision making, for example mining decisions. For many years, it was required for all doctoral students at Harvard Business School.
Description: Systematic approach to ARIMA and ARMAX modelling
Importance: This book introduces ARIMA and associated input-output models, studies how to fit them and develops a methodology for time series forecasting and control. It has changed econometrics, process control and forecasting.
Description: The original manual for researchers, especially biologists, on how to statistically evaluate numerical data.
Importance: Hugely influential text by the father of modern statistics that remained in print for more than 50 years.[5] Responsible for the widespread use of tests of statistical significance.
Description: One of the first comprehensive texts on statistical methods. Reissued as Statistical Methods Applied to Experiments in Agriculture and Biology in 1940 and then again as Statistical Methods with Cochran, WG in 1967. A classic text.
Importance: Influence
Principles and Procedures of Statistics with Special Reference to the Biological Sciences.
Authors: Steel, R.G.D, and Torrie, J. H.
Publication data: McGraw Hill (1960) 481 pages
Description: Excellent introductory text for analysis of variance (one-way, multi-way, factorial, split-plot, and unbalanced designs). Also analysis of co-variance, multiple and partial regression and correlation, non-linear regression, and non-parametric analyses. This book was written before computer programmes were available, so it gives the detail needed to make the calculations manually.Cited in more than 1,381 publications between 1961 and 1975.[6]
Importance: Influence
Biometry: The Principles and Practices of Statistics in Biological Research
Publication data: 1st ed. W. H. Freemann (1969); 2nd ed. W. H. Freemann (1981); 3rd ed. Freeman & Co. (1994)
Description:: Key textbook on Biometry: the application of statistical methods for descriptive, experimental, and analytical study of biological phenomena.
Importance Cited in more than 7,000 publications.[7]
Statistical learning theory
On the uniform convergence of relative frequencies of events to their probabilities
Description: First description of three methods of estimation of variance components in mixed linear models for unbalanced data. "One of the most frequently cited papers in the scientific literature."[9][10]
Description: First description of Minimum Variance Quadratic Unbiased Estimation (MIVQUE) and Minimum Norm Quadratic Unbiased Estimation (MINQUE) for unbalanced data
Description: Generally considered to be the first synthesis of results from separate studies, although no formal statistical methods for combining results are presented.
Importance: Breakthrough, Influence
The Probability Integral Transformation for Testing Goodness of Fit and Combining Independent Tests of Significance
Importance: The first randomized experiment, which also used blinding; it seems also to have been the first experiment for estimating subjective probabilities.[1][2]
Publication data: 1950, John Wiley & Sons, New York (Reprinted with corrections in 1979 by Robert E. Krieger)
Description: Early exposition of the general linear model using matrix algebra (following lecture notes of George W. Brown). Bases inference on the randomization distribution objectively defined by the experimental protocol, rather than a so-called "statistical model" expressing the subjective beliefs of a statistician: The normal model is regarded as a convenient approximation to the randomization-distribution, whose quality is assessed by theorems about moments and simulation experiments.
Importance: The first and most extensive discussion of randomation-based inference in the field of design of experiments until the recent 2-volume work by Hinkelmann and Kempthorne; randomization-based inference is called "design-based" inference in survey sampling of finite populations. Introduced the treatment-unit additivity hypothesis, which was discussed in chapter 2 of David R. Cox's book on experiments (1958) and which has influenced Donald Rubin and Paul Rosenbaum's analysis of observational data.
On the Experimental Attainment of Optimum Conditions (with discussion)
Description: Introduced Box-Wilson central composite design for fitting a quadratic polynomial in several variables to experimental data, when an initial affine model had failed to yield a direction of ascent. The design and analysis is motivated by a problem in chemical engineering.
Importance: Introduced response surface methodology for approximating local optima of systems with noisy observations of responses.
^Pages 560-561 in Sen, Pranab Kumar; Anderson, T. W.; Arnold, S. F.; Eaton, M. L.; Giri, N. C.; Gnanadesikan, R.; Kendall, M. G.; Kshirsagar, A. M.; et al. (June 1986). "Review: Contemporary Textbooks on Multivariate Statistical Analysis: A Panoramic Appraisal and Critique". Journal of the American Statistical Association. 81 (394): 560–564. doi:10.2307/2289251. ISSN0162-1459. JSTOR2289251.
^Stanley, J. C. (1966). "The Influence of Fisher's "The Design of Experiments" on Educational Research Thirty Years Later". American Educational Research Journal. 3 (3): 223–229. doi:10.3102/00028312003003223. S2CID145725524.
