Robust, data-driven inference in non-linear cosmostatistics

Abstract

We discuss two projects in non-linear cosmostatistics applicable to very large surveys of galaxies. The first is a Bayesian reconstruction of galaxy redshifts and their number density distribution from approximate, photometric redshift data. The second focuses on cosmic voids and uses them to construct cosmic spheres that allow reconstructing the expansion history of the Universe using the Alcock-Paczynski test. In both cases we find that non-linearities enable the methods or enhance the results: non-linear gravitational evolution creates voids and our photo-z reconstruction works best in the highest density (and hence most non-linear) portions of our simulations.