Inference Pipeline
This section documents how a set of cosmological parameters becomes a posterior and a model ranking. Everything is implemented in one self-contained script, src/xi_protocol1b_bao_run.py, run once per model on the same data: 1,590 supernovae and 13 BAO measurements.
Pipeline position
| Step | Page | One-line summary |
|---|---|---|
| Priors | Prior Specification | Uniform, bounded; Ξosc amplitude is one-sided |
| Likelihood | Likelihood Functions | Gaussian, full covariance, no nuisance parameters |
| Sampling | MCMC Setup | emcee, 96 walkers, 100,000 steps in the final run |
| Diagnostics | Convergence Diagnostics | N/50 > τmax criterion |
| Ranking | Model Comparison | AIC and BIC from the best sampled point |
Same procedure for every model
ΛCDM, CPL and Ξosc share data, covariances, sampler, walker count, steps and burn-in. Only the expansion history and the parameter vector change.