Marginal Likelihoods from Monte Carlo Markov Chains

Abstract

In this paper, we present a method for computing the marginal likelihood, also known as the model likelihood or Bayesian evidence, from Markov Chain Monte Carlo (MCMC), or other sampled posterior distributions. In order to do this, one needs to be able to estimate the density of points in parameter space, and this can be challenging in high numbers of dimensions. Here we present a Bayesian analysis, where we obtain the posterior for the marginal likelihood, using $k$th nearest-neighbour distances in parameter space, using the Mahalanobis distance metric, under the assumption that the points in the chain (thinned if required) are independent. We generalise the algorithm to apply to importance-sampled chains, where each point is assigned a weight. We illustrate this with an idealised posterior of known form with an analytic marginal likelihood, and show that for chains of length $\sim 10^5$ points, the technique is effective for parameter spaces with up to $\sim 20$ dimensions. We also argue that $k=1$ is the optimal choice, and discuss failure modes for the algorithm. In a companion paper (Heavens et al. 2017) we apply the technique to the main MCMC chains from the 2015 Planck analysis of cosmic background radiation data, to infer that quantitatively the simplest 6-parameter flat $\Lambda$CDM standard model of cosmology is preferred over all extensions considered.

Cited by in corpus (23)

• dynesty: A Dynamic Nested Sampling Package for Estimating Bayesian Posteriors and Evidences
• Curvature tension: evidence for a closed universe
• Interacting dark energy in the early 2020s: a promising solution to the $H_0$ and cosmic shear tensions
• General Cosmography Model with Spatial Curvature
• Tensions in the dark: shedding light on Dark Matter-Dark Energy interactions
• Quantifying tensions in cosmological parameters: Interpreting the DES evidence ratio
• Interacting Dark Energy in a closed universe
• Observational Constraints on Oscillating Dark-Energy Parametrizations
• Cosmological parameter inference with Bayesian statistics
• Partially Acoustic Dark Matter Cosmology and Cosmological Constraints
• Multi-interacting dark energy and its cosmological implications
• Bayesian evidence for the tensor-to-scalar ratio $r$ and neutrino masses $m_{\nu }$: Effects of uniform vs logarithmic priors
• The MUSE-Faint survey. II. The dark matter-density profile of the ultra-faint dwarf galaxy Eridanus 2
• Scale-invariant alternatives to general relativity. III. The inflation--dark-energy connection
• Bayesian Model Selection with Future 21cm Observations of The Epoch of Reionisation
• Probing the Neutrino Mass Hierarchy beyond $\Lambda$CDM Model
• Differentiating dark interactions with perturbation
• Reference level of the vacuum energy density of the Universe and astrophysical data
• Signs of Dynamical Dark Energy in Current Observations
• Interpreting Internal Consistency of DES Measurements
• Dark energy-dark matter interactions as a solution to the $S_8$ tension
• Diagnosing the cosmic coincidence problem and its evolution with recent observations
• Implications of Dynamical Dark Energy in the expansion of the Universe and the Structure Formation