Bayesian nonparametric spectral density estimation using B-spline priors
- 154 Downloads
We present a new Bayesian nonparametric approach to estimating the spectral density of a stationary time series. A nonparametric prior based on a mixture of B-spline distributions is specified and can be regarded as a generalization of the Bernstein polynomial prior of Petrone (Scand J Stat 26:373–393, 1999a; Can J Stat 27:105–126, 1999b) and Choudhuri et al. (J Am Stat Assoc 99(468):1050–1059, 2004). Whittle’s likelihood approximation is used to obtain the pseudo-posterior distribution. This method allows for a data-driven choice of the number of mixture components and the location of knots. Posterior samples are obtained using a Metropolis-within-Gibbs Markov chain Monte Carlo algorithm, and mixing is improved using parallel tempering. We conduct a simulation study to demonstrate that for complicated spectral densities, the B-spline prior provides more accurate Monte Carlo estimates in terms of \(L_1\)-error and uniform coverage probabilities than the Bernstein polynomial prior. We apply the algorithm to annual mean sunspot data to estimate the solar cycle. Finally, we demonstrate the algorithm’s ability to estimate a spectral density with sharp features, using real gravitational wave detector data from LIGO’s sixth science run, recoloured to match the Advanced LIGO target sensitivity.
KeywordsB-spline prior Bernstein polynomial prior Whittle likelihood Spectral density estimation Bayesian nonparametrics LIGO Gravitational waves Sunspot cycle
We thank Claudia Kirch, Alexander Meier, and Thomas Yee for fruitful discussions, and Michael Coughlin for providing us with the recoloured LIGO data. We also thank the New Zealand eScience Infrastructure (NeSI) for their high performance computing facilities, and the Centre for eResearch at the University of Auckland for their technical support. NC’s work is supported by National Science Foundation Grant PHY-1505373. All analysis was conducted in R, an open-source statistical software available on CRAN (cran.r-project.org). We acknowledge the following R packages: Rcpp, Rmpi, bsplinePsd, beyondWhittle, splines, signal, bspec, ggplot2, grid and gridExtra. This paper carries LIGO Document No. LIGO-P1600239.
- Abbott, B.P., et al.: GW170608: observation of a 19-solar-mass binary black hole coalescence. Pre-print, arXiv:1711.05578 (2017d)
- de Boor, C.: B(asic)-spline basics. In: Piegl, L. (ed.) Fundamental Developments of Computer-Aided Geometric Modeling. Academic Press, Washington (1993)Google Scholar
- Edwards, M.C., Meyer, R., Christensen, N.: bsplinePsd: Bayesian power spectral density estimation using B-spline priors. R package (2017)Google Scholar
- Kirch, C., Edwards, M.C., Meier, A., Meyer, R.: Beyond Whittle: nonparametric correction of a parametric likelihood with a focus on Bayesian time series analysis. Pre-print, arXiv:1701.04846v1 (2017)
- Meier, A., Kirch, C., Edwards, M.C., Meyer, R.: beyondWhittle: Bayesian spectral inference for stationary time series. R package (2017)Google Scholar
- Schwabe, S.H.: Sonnenbeobachtungen im jahre 1843. Astron. Nachr. 21, 233–236 (1843)Google Scholar
- Vitale, S., Congedo, G., Dolesi, R., Ferroni, V., Hueller, M., Vetrugno, D., Weber, W.J., Audley, H., Danzmann, K., Diepholz, I., Hewitson, M., Korsakova, N., Ferraioli, L., Gibert, F., Karnesis, N., Nofrarias, M., Inchauspe, H., Plagnol, E., Jennrich, O., McNamara, P.W., Armano, M., Thorpe, J.I., Wass, P.: Data series subtraction with unknown and unmodeled background noise. Phys. Rev. D 90, 042003 (2014)CrossRefGoogle Scholar