Speaker
Description
I present a fully Bayesian MCMC-based signal extraction technique, which also solves the E/B-leakage problem for Stage-IV Surveys caused by their partial sky coverage. For cosmic shear, classical analyses lose information cutting off small scales because the noise dominates its signal, and additionally large scales as a result of the leakage between E and B modes. Our code Almanac allows us to reincorporate those scales and obtain a signal that does not bias the cosmological parameters.
Almanac samples the posterior distribution of al m and Cl given input data as spin−0 or spin−2 fields on the sphere for different redshift bins. The sampling is performed using a guided Hamiltonian Monte Carlo algorithm. In this poster, I produce shear sky maps for two redshift bins under a known wCDM cosmology using a Euclid-like footprint together with a stellar mask. From Almanac, I obtain the full non-Gaussian posterior of the power spectra for E and B modes, and additionally the posterior’s sample-estimated covariance matrix. Additionally, I compute an analytical covariance matrix. With these ingredients, I compare two likelihoods: the full non-Gaussian solution and the standard Gaussian likelihood assumption. From both I infer the wCDM parameters and compare the bias and constraining power of the two approaches.
