Exoplanet masses, probably

Exoplanet masses, probably

Thousands of transiting exoplanets have been discovered, but for most of these planets we only know their radius and nothing about their mass. With a mass-radius relation, we can infer the masses of all these planets. This paper provides a new, probabilistic mass-radius relation for small planets, and its approach is somewhat unusual…

Your Gateway to the Bayesian Realm

This post is written by Benjamin Nelson, a graduate student in the Astronomy Department at the University of Florida. He works with Dr. Eric Ford on the characterization and dynamical evolution of extrasolar planets. He is currently developing an N-body Markov chain Monte Carlo for RV observations of exoplanet systems. Why is this important to astronomy? Inevitably in your astronomical career, you’ll attend some talk where the speaker mentions “MCMC” and “Metropolis-Hastings”, or maybe something about “priors” and “likelihood functions.” The latter terms refer back to a Bayesian framework, while the former terms are the numerical tools, both of which are rarely covered in undergraduate astronomy/physics. Although Bayes’ theorem has been around for more than 200 years, computational advances within only the past couple decades have made it actually practical to solve problems involving Bayesian techniques. Learning statistical methods is like eating your vegetables: you probably won’t enjoy it, but it’ll be good for you in the long run. It is hardly motivating for an astronomy grad student to pick up an introductory book on Bayesian statistics without some practical application in mind, but a solid knowledge of Bayesian methods is a great way to find common ground in other, unfamiliar astronomical subfields, or even other disciplines of science. The purpose of this astrobite is to familiarize the reader with conventional Bayesian jargon (sugar coated with some astronomy) and lay out the ingredients to code a Markov chain Monte Carlo from scratch. Bayes’ Theorem: In short, Bayes’ theorem allows us to update our knowledge of a model system using new sets of observations. We use this to quantify the...
Astrostatistics: How to fit a model to data

Astrostatistics: How to fit a model to data

Why does the fit above look so crappy? Probably because of those pesky outliers! But before you get rid of them, see what David Hogg has to say about alternative methods of fitting models which are not only more robust, but may change your mind about every fit you do from now on.