Abstract

All large-scale structure cosmologists are faced with the question: how do we robustly extract cosmological information, such as on dark energy, gravity, and inflation, from observed tracers such as galaxies whose astrophysics is extremely complex and incompletely understood? I will describe why guaranteeing this robustness is so difficult, and how a perturbative effective-field-theory (EFT) approach offers such a guarantee when focusing on galaxy clustering on large scales. The natural next question then is: how much cosmological information is left on these large scales if we marginalize over all the free parameters introduced in the EFT ? To answer this question, I will introduce our implementation of the EFT on a lattice as an explicit field-level forward model, which can be used both for full Bayesian inference at the field level and for likelihood-free (simulation-based) inference using summary statistics. Finally, I will show first results comparing full field-level inference and summary statistics on fully nonlinear mock tracers.