Likelihood Analysis of Galaxy Surveys

Original · EN

One of the major goals of cosmological observations is to test theories of structure formation. The most straightforward way to carry out such tests is to compute the likelihood function L, the probability of getting the data given the theory. We write down this function for a general galaxy survey. The full likelihood function is very complex, depending on all of the n-point functions of the theory under consideration. Even in the simplest case, where only the two point function is non-vanishing (Gaussian perturbations), L cannot be calculated exactly, primarily because of the Poisson nature of the galaxy distribution. Here we expand L about the (trivial) zero correlation limit. As a first application, we take the binned values of the two point function as free parameters and show that L peaks at (DD - DR + RR)/DD. Using Monte Carlo techniques, we compare this estimator with the traditional DD/DR and Landy & Szalay estimators. More generally, the success of this expansion should pave the way for further applications of the likelihood function.