2025 workshop program
Investigating Survey Systematics with Bayesian Statistics
The work
Abstract
Upcoming Stage IV cosmological surveys will rely on the combined power of weak lensing, galaxy clustering, and galaxy-galaxy lensing (3x2pt) to probe fundamental questions about dark energy and the structure of the universe. It’s statistical power relies on photometric redshifts, which allow us to build a vast and detailed evolutionary history of our universe. However, the uncertainty in the redshifts derived from this method and contamination from intrinsic alignments introduce significant systematic biases. The interacting effects of these systematics can degrade cosmological constraints and bias results, potentially causing us to misidentify the nature of our universe. Using Markov Chain Monte Carlo (MCMC) methods on mock data, I explore the limitations of current model parameterisations and the challenges of identifiability in high-dimensional spaces. Posterior predictive checks are used to evaluate model performance and uncover mismodelling. This work highlights the importance of developing robust, interpretable inference methods that can handle complex, interacting systematics in next-generation surveys.