STATISTICS COLLOQUIUM

THURSDAY, October 15th, 2009 at 4.10pm, MSB 1147 (Colloquium Room)

Refreshments: 3.30pm, MSB 4110 (Statistics Lounge)

Speaker:            Chad M. Shafer (Carnegie Mellon University)

Title:                Testing Cosmological Theories: Methodology for the Inference Challenges

Abstract:        Recent experiments created a glut of astronomical data and a host of interesting problems for statisticians. A primary motivating inference problem is parameter estimation: Physical theories regarding the Universe characterize the distribution of observable data, and these theories are parameterized by relatively few, unknown physical constants. In this talk I focus on methodology we are developing to address the challenges.

For example, WMAP observed the cosmic microwave background radiation (CMB) with high precision, yet estimating cosmological parameters is complicated by the complexity of the stochastic model. In Schafer and Stark (2009) we presented a method that numerically approximates the minimax expected size and minimax regret confidence procedures. This approach to confidence set construction is well suited to CMB inference. Yet, there are further challenges before this (or any similarly-motivated method) can fully exploit the available data. I will discuss directions in addressing some of these issues.

This work is in collaboration with the International Computational Astrostatistics Group and Philip Stark, UC Berkeley.