A Bayesian analysis of the cepheid distance scale

We develop and describe a Bayesian statistical analysis to solve the surface brightness equations for Cepheid distances and stellar properties. Our analysis provides a mathematically rigorous and objective solution to the problem, including immunity from Lutz-Kelker bias. We discuss the choice of priors, show the construction of the likelihood distribution, and give sampling algorithms in a Markov chain Monte Carlo approach for efficiently and completely sampling the posterior probability distribution. Our analysis averages over the probabilities associated with several models rather than attempting to pick the "best model" from several possible models. Using a sample of 13 Cepheids we demonstrate the method. We discuss diagnostics of the analysis and the effects of the astrophysical choices going into the model. We show that we can objectively model the order of Fourier polynomial fits to the light and velocity data. By comparison with theoretical models of Bono et al. we find that EU Tau and SZ Tau are overtone pulsators, most likely without convective overshoot. The period-radius and period-luminosity relations we obtain are shown to be compatible with those in the recent literature. Specifically, we find log(〈R〉) = (0.693 ± 0.037) [log(P) - 1.2] + (2.042 ± 0.047) and 〈Mv〉 = -(2.690 ± 0.169) [log(P) - 1.2] - (4.699 ± 0.216).

Duke Scholars

Author James O. Berger Statistical Science