John W. Tukey's work on time series and spectrum analysis. (Q1873609)

Most of the papers appeared in Volumes I and II of the Collected Works of John Wilder Tukey (1985 and 1986, see Zbl 0699.62084), with some discussion by the editor David R. Brillinger and Tukey's rejoinder. The time series Volumes also contain some noteworthy previously unpublished material, and the other six volumes contain material which touches also time series problems. The present paper begins with a short historical review, beginning in 1664 with Isaac Newton who decomposed a light signal into frequency components, in 1800 J. F. W. Herschel measured the average energy in various bands of the sunlight spectrum, mathematical formulations began to be laid in the mid-1800s, later J. W. Rayleigh replaced the series by an integral, in 1872 Lord Kelvin built a harmonic analyser and a harmonic synthesizer, at the same time G. G. Stockes was carrying out numerical Fourier analyses using computation schedules, in 1898 A. A . Michelson described a harmonic analyser based on springs, 1894--1898 A. Schuster proposed and discussed the periodogram statistics. Starting in 1930 N. Wiener, H. Cramér, A. N. Kolmogorov, M. S. Bartlett, and J. W. Tukey produced substantial developments in time series analysis. This article reviews some of the contributions of J. W. Tukey. Over the years Tukey introduced a multitude of techniques and terms that have become standard to the practice of data analysis generally and of time series specially. In particular we note: prewhitening, aliasing, smoothing and decimation, taper, bispectrum, complex demodulation, cepstrum, saphe craking, quefrency, analysis, ramonic liftering, hamming, hanning, window carpentry, and polyspectrum. Tukey's work has fueled the advance of time series analysis for over fifty years. The work is continuously described as seminal, breakthrough, insightful, essential, and with a host of superlatives. Finally some amusing Tukey stories are being reproduced.