-
Practical Guidance for Bayesian Inference in Astronomy
Authors:
Gwendolyn M. Eadie,
Joshua S. Speagle,
Jessi Cisewski-Kehe,
Daniel Foreman-Mackey,
Daniela Huppenkothen,
David E. Jones,
Aaron Springford,
Hyungsuk Tak
Abstract:
In the last two decades, Bayesian inference has become commonplace in astronomy. At the same time, the choice of algorithms, terminology, notation, and interpretation of Bayesian inference varies from one sub-field of astronomy to the next, which can lead to confusion to both those learning and those familiar with Bayesian statistics. Moreover, the choice varies between the astronomy and statistic…
▽ More
In the last two decades, Bayesian inference has become commonplace in astronomy. At the same time, the choice of algorithms, terminology, notation, and interpretation of Bayesian inference varies from one sub-field of astronomy to the next, which can lead to confusion to both those learning and those familiar with Bayesian statistics. Moreover, the choice varies between the astronomy and statistics literature, too. In this paper, our goal is two-fold: (1) provide a reference that consolidates and clarifies terminology and notation across disciplines, and (2) outline practical guidance for Bayesian inference in astronomy. Highlighting both the astronomy and statistics literature, we cover topics such as notation, specification of the likelihood and prior distributions, inference using the posterior distribution, and posterior predictive checking. It is not our intention to introduce the entire field of Bayesian data analysis -- rather, we present a series of useful practices for astronomers who already have an understanding of the Bayesian "nuts and bolts" and wish to increase their expertise and extend their knowledge. Moreover, as the field of astrostatistics and astroinformatics continues to grow, we hope this paper will serve as both a helpful reference and as a jum** off point for deeper dives into the statistics and astrostatistics literature.
△ Less
Submitted 9 February, 2023;
originally announced February 2023.
-
Clearing the hurdle: The mass of globular cluster systems as a function of host galaxy mass
Authors:
Gwendolyn M. Eadie,
William E. Harris,
Aaron Springford
Abstract:
Current observational evidence suggests that all large galaxies contain globular clusters (GCs), while the smallest galaxies do not. Over what galaxy mass range does the transition from GCs to no GCs occur? We investigate this question using galaxies in the Local Group, nearby dwarf galaxies, and galaxies in the Virgo Cluster Survey. We consider four types of statistical models: (1) logistic regre…
▽ More
Current observational evidence suggests that all large galaxies contain globular clusters (GCs), while the smallest galaxies do not. Over what galaxy mass range does the transition from GCs to no GCs occur? We investigate this question using galaxies in the Local Group, nearby dwarf galaxies, and galaxies in the Virgo Cluster Survey. We consider four types of statistical models: (1) logistic regression to model the probability that a galaxy of stellar mass $M_{\star}$ has any number of GCs; (2) Poisson regression to model the number of GCs versus $M_{\star}$, (3) linear regression to model the relation between GC system mass ($\log{M_{gcs}}$) and host galaxy mass ($\log{M_{\star}}$), and (4) a Bayesian lognormal hurdle model of the GC system mass as a function of galaxy stellar mass for the entire data sample. From the logistic regression, we find that the 50% probability point for a galaxy to contain GCs is $M_{\star}=10^{6.8}M_{\odot}$. From post-fit diagnostics, we find that Poisson regression is an inappropriate description of the data. Ultimately, we find that the Bayesian lognormal hurdle model, which is able to describe how the mass of the GC system varies with $M_{\star}$ even in the presence of many galaxies with no GCs, is the most appropriate model over the range of our data. In an Appendix, we also present photometry for the little-known GC in the Local Group dwarf Ursa Major II.
△ Less
Submitted 28 October, 2021;
originally announced October 2021.
-
Introducing Bayesian Analysis with $\text{m&m's}^\circledR$: an active-learning exercise for undergraduates
Authors:
Gwendolyn Eadie,
Daniela Huppenkothen,
Aaron Springford,
Tyler McCormick
Abstract:
We present an active-learning strategy for undergraduates that applies Bayesian analysis to candy-covered chocolate $\text{m&m's}^\circledR$. The exercise is best suited for small class sizes and tutorial settings, after students have been introduced to the concepts of Bayesian statistics. The exercise takes advantage of the non-uniform distribution of $\text{m&m's}^\circledR~$ colours, and the di…
▽ More
We present an active-learning strategy for undergraduates that applies Bayesian analysis to candy-covered chocolate $\text{m&m's}^\circledR$. The exercise is best suited for small class sizes and tutorial settings, after students have been introduced to the concepts of Bayesian statistics. The exercise takes advantage of the non-uniform distribution of $\text{m&m's}^\circledR~$ colours, and the difference in distributions made at two different factories. In this paper, we provide the intended learning outcomes, lesson plan and step-by-step guide for instruction, and open-source teaching materials. We also suggest an extension to the exercise for the graduate-level, which incorporates hierarchical Bayesian analysis.
△ Less
Submitted 16 April, 2019;
originally announced April 2019.
-
Bayesian Mass Estimates of the Milky Way: including measurement uncertainties with hierarchical Bayes
Authors:
Gwendolyn Eadie,
Aaron Springford,
William Harris
Abstract:
We present a hierarchical Bayesian method for estimating the total mass and mass profile of the Milky Way Galaxy. The new hierarchical Bayesian approach further improves the framework presented by Eadie, Harris, & Widrow (2015) and Eadie & Harris (2016) and builds upon the preliminary reports by Eadie et al (2015a,c). The method uses a distribution function $f(\mathcal{E},L)$ to model the galaxy a…
▽ More
We present a hierarchical Bayesian method for estimating the total mass and mass profile of the Milky Way Galaxy. The new hierarchical Bayesian approach further improves the framework presented by Eadie, Harris, & Widrow (2015) and Eadie & Harris (2016) and builds upon the preliminary reports by Eadie et al (2015a,c). The method uses a distribution function $f(\mathcal{E},L)$ to model the galaxy and kinematic data from satellite objects such as globular clusters (GCs) to trace the Galaxy's gravitational potential. A major advantage of the method is that it not only includes complete and incomplete data simultaneously in the analysis, but also incorporates measurement uncertainties in a coherent and meaningful way. We first test the hierarchical Bayesian framework, which includes measurement uncertainties, using the same data and power-law model assumed in Eadie & Harris (2016), and find the results are similar but more strongly constrained. Next, we take advantage of the new statistical framework and incorporate all possible GC data, finding a cumulative mass profile with Bayesian credible regions. This profile implies a mass within $125$kpc of $4.8\times10^{11}M_{\odot}$ with a 95\% Bayesian credible region of $(4.0-5.8)\times10^{11}M_{\odot}$. Our results also provide estimates of the true specific energies of all the GCs. By comparing these estimated energies to the measured energies of GCs with complete velocity measurements, we observe that (the few) remote tracers with complete measurements may play a large role in determining a total mass estimate of the Galaxy. Thus, our study stresses the need for more remote tracers with complete velocity measurements.
△ Less
Submitted 16 January, 2017; v1 submitted 20 September, 2016;
originally announced September 2016.
