Yichen Tang
Title: Two-Stage Bayesian Analysis of Thick-Disk Galactic White Dwarfs
Date: Thursday, May 22nd
Time: 2:00pm
Location: LIB 2020
Supervised by: David Stenning
Abstract: Inferring the ages of thick-disk white dwarf stars helps astrophysicists explore theories of galaxy evolution. To aid in this goal, "black box" software for performing Bayesian analyses of individual white dwarfs is available, but it is not easily editable to fit more complex, hierarchical statistical models. Instead, we adapt an existing two-stage approach that uses the black-box software in an initial stage to obtain a posterior sample for each white dwarf, which is then carried forward for use as a proposal distribution for a Metropolis-Hastings algorithm in the second stage. This allows a hierarchical model for thick-disk white dwarf ages to be fit without re-evaluating the likelihood function. The two-stage algorithm is demonstrated on a normal-normal hierarchical model, and we further explore our model and methods through a series of simulation studies. Finally, we analyze 30 thick-disk white dwarfs and provide posterior summaries that will be useful to the astrophysics community.
Keywords: Bayesian methodology; stellar evolution; (thick-disk) white dwarfs; Markov chain Monte Carlo; hierarchical modeling; two-stage algorithms