Selected papers of Hirotugu Akaike (Q1271115)

The selected papers in this book are divided into six groups, representing successive phases of Akaike's research interests during his more than 40 years of work at the prestigious Institute of Statistical Mathematics in Tokyo (from 1952 until his retirement in 1994). He was Director General of the Institute from 1986 to 1994. The Institute provided Akaike with a unique environment which he felt protected his freedom of choice of subject and way of developing research. 1. Two papers, called Precursors, represent the years 1952-1960 which were a launching period; a paper on gap (zero-one valued) processes and a paper on methods of optimization and numerical analysis. 2. Three papers are reprinted from the years 1960-1965, emphasizing frequency domain time series analysis; they are concerned with power-spectrum analysis and frequency response analysis. 3. Six papers describe Akaike's innovative research after 1965 emphasizing time domain time series analysis with influential (and prize winning) applications of these methods to analysis of feedback systems and autoregressive model fitting for control of cement rotary kilns and power plants. 4. Six papers describe Akaike's research after 1970 which developed his world famous criterion AIC for model identification. Fundamental research on using state space models to identify parsimonious models for multivariate time series is presented in the 1974 paper on Markovian representation of stochastic processes [Ann. Inst. Statist. Math. 26, 363-387 (1974; Zbl 0335.62058)]. 5. Nine papers represent Akaike's research after 1977 on Bayesian statistical analysis and entropy methods of statistical inference whose goal is the practical application of Bayesian models by developing criteria to compare competing models and priors. 6. Three papers written since 1985 exposit Akaike's philosophy of statistical thinking. The review paper ``Prediction and entropy'', published in `A celebration of statistics', The ISI Centen. Vol., 1-24 (1985; Zbl 0576.62009), summarizes the statistical literacy requirements of a modern statistician who wants to benefit from Akaike's research strategy: AIC, Bayes procedure, entropy, entropy maximization principle, information, likelihood, model selection, predictive distribution. Many researchers at the Institute of Statistical Mathematics and throughout the world profitably apply information methods of statistical inference developed by Akaike as the result of many years of successfully applying statistics to important real problems. We hope that these ``Selected papers of Hirotugu Akaike'' will stimulate future generations of statisticians and applied researchers to learn and teach Akaike's philosophy of statistical research.