Suite of optimal and efficient power spectrum and bispectrum estimators for large-scale structure analyses

By measuring, modeling and interpreting cosmological datasets, one can place strong constraints on models of the Universe. Central to this effort are summary statistics such as power spectra and bispectra, which condense the high-dimensional data into low-dimensional representations. In this work, we introduce a modern set of estimators for computing such statistics from three-dimensional clustering data, and provide a flexible Python/Cython implementation; PolyBin3D. Working in a maximum-likelihood formalism, we derive general estimators for the two- and three-point functions, which yield unbiased spectra regardless of the survey mask and weighting scheme. These can be directly compared to theory without the need for mask-convolution. Furthermore, we present a numerical scheme for computing the optimal (minimum-variance) estimators for a given survey, which is shown to reduce error-bars on large-scales. Our Python package includes both general “unwindowed” estimators and their idealized equivalents (appropriate for simulations), each of which are efficiently implemented using fast Fourier transforms and Monte Carlo summation tricks, and additionally supports graphics processing unit acceleration using JAX. These are extensively validated in this work, with Monte Carlo convergence (relevant for masked data) achieved using only a small number of iterations (typically < 10 for bispectra). This will allow for fast and unified measurement of two- and three-point functions from current and upcoming survey data.