Outputs and plots

Chain files

With output="chains/run", every sampler writes Cobaya-format files that getdist and anesthetic read as they are:

File	Content
run.1.txt, ...	One chain per file: weight, \(-\log\) posterior, parameters, derived (*)
run.covmat	Covariance of the sampled parameters
run.updated.yaml	Settings, labels and the results diagnostics
run.adaptation.npz	Step size and mass matrix, for resuming NUTS and HMC
run_dead-birth.txt, run.nested_state.npz	Nested sampling's points and resume state
run.bestfit.yaml, run.profile.*	From maximize and profile

save_chains writes the same format from any result.

from getdist import loadMCSamples

samples = loadMCSamples("chains/run")

Diagnostics

NUTS and HMC results carry \(\hat R\), ESS and divergences; result.summary() prints them. rhat works on any (chains, n, d) array, with the "rank" (default) or "gelman-rubin" variant.

Figures

All figures use cplots, installed through the plots dependency group:

uv sync --group plots
uv run --group plots cosmonuts plot triangle chains/run -p H0,Omega_m,w0,wa
uv run --group plots cosmonuts plot trace chains/run
uv run --group plots cosmonuts plot stats chains/run --limits

plot triangle overlays several runs (roots or run files), marks truths with -t name=value and sizes for a journal with --paper PRD|PRD-2col|MNRAS|A4. From Python:

import cplots
from cosmonuts.plotting import load_run

run = load_run("chains/run")
cplots.TrianglePlot(run.samples, run.parameter_names, weights=run.weights).save("triangle.pdf")

Note

Do not build figures with matplotlib.pyplot or getdist directly in this project: cplots wraps both.