An approach to robust Bayesian regression in astronomy

Abstract

Model mis-specification (e.g. the presence of outliers) is commonly encountered in astronomical analyses, often requiring the use of ad hoc algorithms which are sensitive to arbitrary thresholds (e.g. sigma-clipping). For any given dataset, the optimal approach will be to develop a bespoke statistical model of the data generation and measurement processes, but these come with a development cost; there is hence utility in having generic modelling approaches that are both principled and robust to model mis-specification. Here we develop and implement a generic Bayesian approach to linear regression, based on Student's t-distributions, that is robust to outliers and mis-specification of the noise model. Our method is validated using simulated datasets with various degrees of model mis-specification; the derived constraints are shown to be systematically less biased than those from a similar model using normal distributions. We demonstrate that, for a dataset without outliers, a worst-case inference using t-distributions would give unbiased results with $\lesssim 10$ per cent increase in the reported parameter uncertainties. We also compare with existing analyses of real-world datasets, finding qualitatively different results where normal distributions have been used and agreement where more robust methods have been applied. A Python implementation of this model, t-cup, is made available for others to use.

16 pages, 21 figures; accepted for publication in RASTI

References in corpus (28)

• A 3% Solution: Determination of the Hubble Constant with the Hubble Space Telescope and Wide Field Camera 3
• Some Aspects of Measurement Error in Linear Regression of Astronomical Data
• Revisiting the Scaling Relations of Black Hole Masses and Host Galaxy Properties
• The Galaxy Luminosity Function and Luminosity Density at Redshift z=0.1
• The Low-Luminosity End of the Radius-Luminosity Relationship for Active Galactic Nuclei
• The Mass Dependence Between Protoplanetary Disks and their Stellar Hosts
• H0 Revisited
• Extreme deconvolution: Inferring complete distribution functions from noisy, heterogeneous and incomplete observations
• Clarifying the Hubble constant tension with a Bayesian hierarchical model of the local distance ladder
• Secure TTV Mass Measurements: Ten Kepler Exoplanets between 3 and 8 Earth Masses with Diverse Densities and Incident Fluxes
• Shedding Light on the Galaxy Luminosity Function
• Investigating hot-Jupiter inflated radii with hierarchical Bayesian modelling
• A Flexible Method of Estimating Luminosity Functions
• Hierarchical Bayesian inference of galaxy redshift distributions from photometric surveys
• Determining Quasar Black Hole Mass Functions from their Broad Emission Lines: Application to the Bright Quasar Survey
• Galaxy evolution in the high redshift, colour-selected cluster RzCS 052 at z=1.02
• Extending the Calibration of CIV-Based Single-Epoch Black Hole Mass Estimators for Active Galactic Nuclei
• Not Normal: the uncertainties of scientific measurements
• Type Ia Supernovae are Excellent Standard Candles in the Near-Infrared
• The scaling relation between richness and mass of galaxy clusters: a Bayesian approach
• A New Bayesian Approach to Robustness Against Outliers in Linear Regression
• A Gibbs Sampler for Multivariate Linear Regression
• 21st Century Statistical and Computational Challenges in Astrophysics
• Robust and Accurate Inference via a Mixture of Gaussian and Student's t Errors
• Marginalised Normal Regression: Unbiased curve fitting in the presence of x-errors
• Evidence for radially independent size growth of early-type galaxies in clusters
• Theoretical properties of Bayesian Student-$t$ linear regression
• Posterior Robustness with Milder Conditions: Contamination Models Revisited

Cited by in corpus (2)

• The Limits of Photometric Dynamics: Benchmarking Cluster Relaxation Diagnostics
• Cold-Atom Buoy: A Differential Magnetic Sensing Technique in Cold Quadrupole Traps