Parameter estimation and model selection in extragalactic astronomy

Abstract

Astronomy is rife with multi-instrument, multiple wave band data sets and complex physical theories. An astronomer, therefore, needs to (1) infer the parameters of models from multiple hypotheses; (2) inter-compare hypotheses; and (3) test that the data is sufficiently well explained by the models. Most often, all three needs are inseparably linked. The Bayesian approach allows these to be addressed simultaneously and consistently. Although Bayesian inference is wellsuited to problems of inference in astronomical science, the most commonly used tools best treat idealized or specialized models. Here, I describe our experience based on two such problems in extragalactic science-testing models based on galaxy images and exploring recipes galaxy evolution using semi-analytic models- using the UMass Bayesian Inference Engine (BIE), a parallel-optimized software package for parameter inference and model selection. The BIE is designed as a collaborative platform for Bayesian methodology for astronomical problems. © Springer Science+Business Media New York 2013.