Towards optimal cosmological parameter recovery from compressed bispectrum statistics

Over the next decade, improvements in cosmological parameter constraints will be driven by surveys of a large-scale structure in the Universe. The information they contain can be measured by suitably chosen correlation functions, and the non-linearity of structure formation implies that significant information will be carried by the 3-point function or higher correlators. Extracting this information is extremely challenging, requiring accurate modelling and significant computational resources to estimate the covariance matrix describing correlation between different Fourier configurations. We investigate whether it is possible to reduce this matrix without significant loss of information by using a proxy that aggregates the bispectrum over a subset of configurations. Specifically, we study constraints on Lambda CDM parameters from a future galaxy survey combining the power spectrum with (a) the integrated bispectrum, (b) the line correlation function and (c) the modal decomposition of the bispectrum. We include a simple estimate for the degradation of the bispectrum with shot noise. Our results demonstrate that the modal bispectrum has comparable performance to the Fourier bispectrum, even using considerably fewer modes than Fourier configurations. The line correlation function has good performance, but is less effective. The integrated bispectrum is comparatively insensitive to the background cosmology. Addition of bispectrum data can improve constraints on bias parameters and sigma(8) by a factor between 3 and 5 compared to power spectrum measurements alone. For other parameters, improvements of up to similar to 20 per cent are possible. Finally, we use a range of theoretical models to explore the sophistication required to produce realistic predictions for each proxy.