Bayesian Model Selection and Parameter Estimation Applied to Sea Floor Pressure Data
A Bayesian model and parameter estimation algorithm is described which determines both the model order and the model parameters. The Bayesian approach also allows the removal of parameters, which are of no interest, from the models, thus concentrating attention on only those parameters which are of interest. Artificial time series are analyzed with the algorithm, which is shown to be robust in the presence of noise, and the results are compared with those obtained using a Fourier transform technique. An original time series consisting of 383 hourly measurements of the sea floor pressure beneath the Bellinghausen Sea, Antarctica is also examined. The frequency components of the pressure variations are known accurately from related calculations, but the amplitudes and phases are not well known. The Bayesian algorithm is used to determine both the number of sinusoidal components and their individual frequencies, amplitudes and phases. It is concluded that the Bayesian approach is superior, to the commonly used Fourier transform approach, in the identification of the frequency components of both real and artificial time series.
KeywordsBayesian Approach Prior Information Original Time Series Bayesian Algorithm BAYESIAN Model Selection
Unable to display preview. Download preview PDF.
- 1.Bretthorst, G. L. (1989) ‘Bayesian Spectrum Analysis and Parameter Estimation’, Springer-Verlag.Google Scholar
- 2.Jaynes, E. T. (1989) ‘Information Theory and Statistical Mechanics’, in R. D. Rosenkrantz (ed.). E. T. Jaynes: Papers on Probability, Statistics and Statistical Physics, Kluwer Academic Publishers, Dordrecht, pp 6–16.Google Scholar
- 3.Melchior, P. (1983) ‘The Tides of the Planet Earth’, Pergammon Press, Oxford.Google Scholar