Improving Stacking Techniques with Bayesian
Speaker: Michelle Knights
Venue/Time: SAAO Auditorium/16h00
Abstract: Stacking is a technique used in almost every field in astronomy whereby multiple noisy datasets (for example spectra or images) are co-added to improve the total signal-to-noise and allow average signals to be pulled out of the data. However, traditional stacking techniques can be inflexible (often requiring precise position data, for example) and lose information that could potentially be extracted. In this talk, I will introduce a new approach to stacking using Bayesian statistics, explaining in detail how one can potentially use hierarchical modelling to obtain information about the underlying distribution from which the data are drawn. I will illustrate the idea with some simple examples and explain the Multiple Block Metropolis Hastings algorithm I have implemented to solve the computational challenges that arise with this method.