API reference¶

Everything importable from CosmoFit itself is on this page; the subpackages below carry the rest.

Models¶

Twenty expansion histories written out by hand, and three routes to one that is not here. Every one is a Cosmology subclass, so every dataset, likelihood, sampler and plot works on all of them without knowing which.

`Cosmology`	Base class for all cosmological models.
`CosmologyParameters`	Generic container for the parameters shared by every cosmological model implemented in CosmoFit (LCDM, CPL, ...).
`ModelConfigurationError`	The model cannot answer this question at any parameter values.
`LCDM`	Flat or curved Lambda-CDM.
`WCDM`	Flat or curved wCDM: dark energy with a constant equation of state w0 (does not evolve with redshift; `wa` is ignored, same as `LCDM` ignoring `w0`/`wa` and `CPL` being the two-parameter, `wa`-evolving generalization of this model).
`CPL`	Chevallier-Polarski-Linder (CPL) dark-energy model.
`JBP`	Jassal-Bagla-Padmanabhan (JBP) dark-energy model.
`BA`	Barboza-Alcaniz (BA) dark-energy model.
`LogarithmicDE`	Logarithmic (Efstathiou) dark-energy parametrization,
`PEDE`	Phenomenologically Emergent Dark Energy (Li & Shafieloo 2019).
`GEDE`	Generalized Emergent Dark Energy (Li & Shafieloo 2020).
`LsCDM`	Lambda_s CDM (Akarsu et al. 2021): LCDM, except that the cosmological constant changes sign at a redshift `z_dagger`.
`GCG`	Generalized Chaplygin Gas (GCG).
`IDE`	Interacting Dark Energy: dark matter and dark energy exchange energy, with a coupling `Q = 3 xi H rho_de`.
`RunningVacuum`	Running Vacuum Model: a cosmological "constant" that runs with the expansion rate,
`Cardassian`	Modified Polytropic Cardassian expansion (Freese & Lewis 2002; Wang et al. 2003),.
`HDE`	Holographic dark energy with a future-event-horizon cutoff.
`ADE`	New agegraphic dark energy: the holographic cutoff is the age of the universe, in conformal time.
`RDE`	Ricci dark energy: the holographic cutoff is the Ricci scalar.
`DGP`	Dvali-Gabadadze-Porrati braneworld gravity, self-accelerating branch.
`FQExponential`	f(Q) gravity, exponential model.
`FRTLinear`	f(R,T) gravity, linear model: f(R,T) = R + 2lambdaT.
`FRHuSawicki`	f(R) gravity, Hu & Sawicki (2007) model.
`define_model`	Build a new `Cosmology` subclass from a plain `E(z)` function.
`model_from_expression`	Same as `define_model()`, but `E`/`w`/`dEdz`/ `Omega_de`/`mu` are given as Python expression strings (e.g. `"sqrt(Omega_m(1+z)3 + (1-Omega_m)(1+z)*(3(1+w0)))"`) instead of callables -- convenient for text-entry UIs (see the Streamlit app under `app/`), where asking for a Python function isn't practical.

Likelihoods¶

One class per dataset, plus the joint likelihood that combines them. Three of these are tabulated rather than Gaussian – they carry a released likelihood surface instead of a mean and a covariance.

`BaseLikelihood`
`JointLikelihood`	Joint likelihood composed of multiple independent likelihoods.
`CCLikelihood`
`DESILikelihood`
`SDSSBAOLikelihood`	Combined SDSS BAO-only likelihood: BOSS DR12 (z=0.38, 0.51) + eBOSS DR16 LRG (z=0.698) + eBOSS DR16 QSO (z=1.48), each an independent, non-overlapping-redshift measurement -- see `load_sdss_bao()` for why BOSS DR12's usual third bin (z=0.61) is omitted.
`SDSSFullShapeLikelihood`	SDSS BAO + full-shape consensus: `D_M/r_d`, `D_H/r_d` and `f sigma_8` at z = 0.38, 0.51, 0.698 and 1.48, twelve numbers with the covariance between them.
`EBOSSELGLikelihood`	eBOSS DR16 emission-line galaxies, `D_V/r_d` at `z_eff = 0.845`, as a tabulated likelihood.
`EBOSSELGFullShapeLikelihood`	eBOSS DR16 emission-line galaxies, full shape: the joint likelihood of `(D_M/r_d, D_H/r_d, f sigma_8)` at `z_eff = 0.845`, on a 100x100x100 grid.
`EBOSSLyaLikelihood`	eBOSS DR16 Lyman-alpha forest BAO, `(D_M/r_d, D_H/r_d)` at `z_eff = 2.334`, as a tabulated two-dimensional likelihood.
`BAOLowZLikelihood`	The two BAO measurements below z = 0.2: 6dFGS at z = 0.106 (Beutler et al. 2011) and the SDSS DR7 Main Galaxy Sample at z = 0.15 (Ross et al. 2015).
`PantheonLikelihood`	Pantheon+ / Pantheon+SH0ES Supernova likelihood.
`DESSN5YRLikelihood`	DES-SN5YR (Dark Energy Survey 5-year) Supernova likelihood.
`Union3Likelihood`	Union3 (Rubin et al. 2023): 2087 supernovae compressed into 22 binned distance moduli with a full 22x22 covariance.
`PlanckLikelihood`
`PlanckLiteLikelihood`	Planck 2018 `plik_lite` binned TT/TE/EE bandpowers against a Boltzmann-computed C_l spectrum.
`PlanckLensingLikelihood`	Planck 2018 lensing-potential bandpowers against a Boltzmann-computed `C_L^{phiphi}`.
`PlanckLowEELikelihood`	Planck 2018 low-multipole EE, via its released probability table.
`ACTDR6LensingLikelihood`	ACT DR6 lensing bandpowers against a Boltzmann-computed `C_L^{kappakappa}`.
`FSigma8Likelihood`	Growth-rate (fsigma8) likelihood, e.g. the "Gold-2018" RSD compilation (`load_fsigma8()`).
`S8Likelihood`	A single Gaussian S8 = sigma8 * sqrt(Omega_m / 0.3) constraint from a weak-lensing survey (default: KiDS-1000; also DES Y3 -- see `load_s8()`).
`H0Likelihood`	A local distance-ladder H0 measurement.
`OmegaBLikelihood`	A BBN constraint on the physical baryon density `omega_b h^2`.
`TauLikelihood`	The Planck 2018 large-scale-polarization constraint on the reionization optical depth, `tau = 0.0544 +- 0.0073`.

Fitting¶

`Fitter`	Ties together a cosmological model, one or more datasets, a prior, and (optionally) an MCMC run and best-fit search.
`BaseSampler`	Interface a sampling backend must implement to be usable by `Fitter`.
`EnsembleSampler`	`emcee.EnsembleSampler` backend: affine-invariant ensemble MCMC, the default (and currently only) way `Fitter.run_mcmc` explores a posterior.
`FitResult`	Everything a finished (or in-progress) fit produced, in one object: which model/datasets/parameters it is, the best-fit point (if `best_fit()` was called), and the MCMC posterior (if `run_mcmc()` was called).
`BestFitResult`	Maximum-likelihood point, as found by `Fitter.best_fit()`.
`MCMCResult`	Posterior summary of an MCMC run, as found by `Fitter.run_mcmc()`.

Plotting¶

FitPlotter

Plotting methods for a Fitter.

Saved chains¶

A chain is written as it is sampled, resumed rather than re-sampled next time, and can be reopened months later without loading a single dataset.

`ChainFile`	An HDF5 file holding one MCMC chain, plus the CosmoFit metadata describing it.
`StoredSampler`	A chain read back from disk, wearing the small part of the `emcee.EnsembleSampler` interface that CosmoFit actually asks of it.
`open_chain`	Read a saved chain back for analysis, without a `Fitter` and without touching any dataset.
`chain_info`	What's in a chain file -- model, datasets, parameters, how many steps -- without reading the chain itself.

Datasets¶

`available_datasets`
`available_versions`	Return all available versions of a dataset.
`dataset_reference`	The citation string for a dataset version, without loading any of its files.

Types¶

The two array types this library’s public surface speaks in.

Every model here takes a redshift and returns a number per redshift, and both ends are deliberately loose about shape: you may pass a Python float, a list, or an array, and what comes back matches.

Spelling that out once is what makes the rest of the annotations honest. E(0.5) returns np.float64 and E([0.1, 0.2]) returns an ndarray – np.asarray(z, dtype=float) on a scalar gives a 0-d array, and numpy’s ufuncs turn that back into a scalar on the way out. Annotating the return as np.ndarray alone would be wrong for the commonest call in the library.

>>> from CosmoFit.typing import Array, Redshift
>>>
>>> def E(self, z: Redshift) -> Array:
...     ...

CosmoFit.typing.Array = numpy.float64 | NDArray[numpy.float64]

a scalar for a scalar, an array for an array.

Type:: What comes back

CosmoFit.typing.PathLike = str | os.PathLike

a string, or anything implementing the os.PathLike protocol (pathlib.Path, most obviously).

Type:: Anywhere a file is written or read

type CosmoFit.typing.Redshift = Buffer | _DualArrayLike[dtype, complex | bytes | str]: A redshift, scale factor or wavenumber: anything numpy can make an array of floats from – a scalar, a sequence, or an array.