We are organizing a workshop focusing on statistical methods in cosmology, Jan 11-13 at Berkeley, hosted by the Berkeley Center for Cosmological Physics. The main subject is sampling and non-sampling methods in cosmology. The purpose of the workshop is to bring together experts on cosmological data analysis, with the focus on the sampling methods in cosmology, and their alternatives. Some of the topics for discussion will be the current state of the art statistical and data analysis methods at various stages of data processing, from image processing or time stream reductions to low order statistics, to cosmological parameters. We also plan to discuss methods of systematics cleaning (foregrounds, extinction etc.), joint map and cosmology analyses (Gibbs and Hamltonian samplers and alternatives) etc. Part of the discussion we hope to have is when are sampling preferable over non-sampling methods. In cosmic microwave background analysis one does full likelihood analysis, optimal quadratic estimation, Pseudo-Cl (all non-sampling methods) or Gibbs sampling to get Cl's. In large-scale structure, current analyses focus on two point functions, while sampling methods are just starting to become relevant, with Hamiltonian sampling the current choice. When projecting from Cl's to cosmological parameters, various sampling methods (e.g. MCMC) are the preferred choice. Our primary goal is to have discussions and presentations of different practices based on various practical applications. We are planning 1 day CMB, 1 day large-scale structure and 1 day other topics and future directions.
There will be no registration fee, and all coffee breaks will be sponsored by BCCP and Peder Sather foundation.
Registration is now closed. If you would still like to come please email Marcel Schmittfull.