Bayesian method for frequency detection in paleoclimatic proxy data

The plausibility of frequencies in paleoclimatic proxy data is a longstanding question. In this study radiocarbon calibration data is analysed. In the past, various analyses of the radiocarbon data have been made and some spectral peaks found. However, condence levels, if used at all, have been rather arbitrary. Other problems include the inability to deal with invidual uncertainties, effect of trend removal on the results, and the assumption of white noise. These are common problems in proxy data analysis. In this study, a Bayesian multiple-frequency model is presented which uses variable rst-order correlated noise, enabling the noise color to vary from white to red. The model alleviates the above mentioned problems and deals naturally with unevenly spaced data. Computation is done with a reversible jump markov chain monte carlo algorithm. The result of the analysis with the present method is, somewhat contrary to the previous analyses, that there are no believable continous sinusoids in the radiocarbon calibration data