Bayesian spectrum analysis and parameter estimation. (Extensive revision of the dissertation, Washington Univ. 1987) (Q1188541)

In this work, which is an extensive revision of his Ph. D. dissertation, the author develops and applies some of the aspects of Bayesian parameter estimation to time series, even though the analysis as formulated is applicable to any data set. Besides the original author's results a great deal of introductory and tutorial material is included. The main text is divided into eight chapters. Chapter 1 contains introduction, historical perspective and methods of calculation. In Chapter 2, a time series with a simple stationary harmonic signal plus noise is analyzed. In Chapter 3, these simple considerations are generalized to complex models including the problem of estimating the spectrum of multiple nonstationary harmonic frequencies in the presence of noise. The parameters of the time series are estimated and calculated in Chapter 4. A procedure for assessing the accuracy of the estimates is also derived. In Chapter 5, the problem of selecting the ``best'' model of a process is considered. Chapter 6 is devoted to a specialized discussion on spectrum analysis. Chapter 7 contains applications of the theory to a number of real time series, and Chapter 8 contains summary and conclusions.