Speaker
Description
DESI DR1 full-shape clustering constrains cosmology from linear to quasi-nonlinear scales but is vulnerable to projection effects from broad nuisance priors. We strengthen the analysis along three axes: (i) HOD-informed priors (HIP), calibrated on high-fidelity mocks, to anchor nuisance parameters to realistic galaxy–halo connections; (ii) nonlinear reparameterization of the nuisance sector (via Generalized Additive Models) to reduce degeneracies and likelihood curvature; and (ii i) a prior-free frequentist pipeline based on profile-likelihood scans that guarantees correct coverage. Applied to the same DESI DR1 data, HIP and reparameterization mitigate projection effects and sharpen constraints, while the frequentist scans provide an independent cross-check, and offers a distinct, likelihood-based interpretation of the data. All three approaches yield consistent results, establishing a robust, projection-resilient DESI full-shape cosmology framework.
