|
Activity Number:
|
411
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 1, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308841 |
|
Title:
|
Fully Bayesian Analysis of Low-Count Astronomical Images
|
|
Author(s):
|
David van Dyk*+ and Alanna Connors
|
|
Companies:
|
University of California, Irvine and Eurika Scientific
|
|
Address:
|
2206 Bren Hall, Irvine, CA, 92697-1250,
|
|
Keywords:
|
Astrostatistics ; Multi-Scale Methods ; Image Analysis ; MCMC in Practice ; Bayesian Methods ; Poisson Models
|
|
Abstract:
|
New space-based telescopes that are designed to map X-ray and gamma-ray emission are giving a completely new perspective on the hot and turbulent regions of the universe. Analysis of the resulting images is a sophisticated task that requires subtle statistical techniques. Data is collected as photon counts on a grid of detector pixels. The counts are subject to non-uniform stochastic censoring, heteroscedastic errors in measurement, and background contamination. This combined with relatively small datasets makes answering complex astronomical questions a challenge. In this talk I describe how we (1) use Markov-random-field or multi-scale priors to stabilize the fitted images; (2) use posterior simulation to quantify uncertainty; (3) use higher resolution radio data to inform our priors; and (4) use Bayesian methods to test for deviations from particular structures in the image.
|
