Source code for cosmofit.likelihoods.sdss_bao
"""
SDSS BAO likelihood (BOSS DR12 + eBOSS DR16 LRG + eBOSS DR16 QSO).
"""
from __future__ import annotations
from cosmofit.data.loader import load_sdss_bao
from cosmofit.data.loader import load_sdss_fsbao
from .desi import BAODistanceLikelihood
[docs]
class SDSSBAOLikelihood(BAODistanceLikelihood):
"""
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
:func:`~data.loader.load_sdss_bao` for why BOSS DR12's usual
third bin (z=0.61) is omitted.
Same (z, value, observable-type) structure as
:class:`~likelihoods.desi.DESILikelihood`, so it shares the
latter's ``model()``/``residuals()``/``chi2()`` implementation
via :class:`~likelihoods.desi.BAODistanceLikelihood`.
Warning
-------
Do not combine ``"desi"`` and ``"sdss_bao"`` in the same
``Fitter`` fit -- DESI's own footprint and redshift ranges
overlap the BOSS/eBOSS samples this dataset is built from (DESI
is, among other things, an SDSS successor survey targeting much
of the same sky), so treating them as independent would
double-count structure.
References
----------
Alam et al. (2017), arXiv:1607.03155 (BOSS DR12).
eBOSS Collaboration / Alam et al. (2021), arXiv:2007.08991
(eBOSS DR16 summary).
Data as distributed by CobayaSampler/bao_data (originally with
CosmoMC).
"""
[docs]
def __init__(
self,
cosmology,
version="dr12+dr16",
):
dataset = load_sdss_bao(
version,
)
super().__init__(
name="SDSS-BAO",
dataset=dataset,
cosmology=cosmology,
)
[docs]
class SDSSFullShapeLikelihood(BAODistanceLikelihood):
"""
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.
The same galaxies as :class:`SDSSBAOLikelihood`, analysed for
their full anisotropic clustering rather than the BAO peak
alone. Two things follow. It measures the **growth rate**, so
unlike every other BAO dataset here it constrains ``sigma8``.
And it carries the **correlation between growth and geometry**
-- 0.19 to 0.64 between ``D_M/r_d`` and ``f sigma_8`` within a
bin, strongest for the quasars -- which is exactly what is thrown away by using the
BAO-only dataset together with a separate ``f sigma_8``
compilation drawn from the same surveys.
``f sigma_8`` is compared with the model's own ``fsigma8(z)``
with **no Alcock-Paczynski rescaling**, unlike
:class:`~likelihoods.fsigma8.FSigma8Likelihood`: a full-shape
fit varies the geometry alongside the growth rate, so the
fiducial is already fitted rather than something to correct
back to.
Warning
-------
Mutually exclusive with ``"sdss_bao"`` (the same BAO
measurements) and with ``"fsigma8"`` (whose compilation
includes these surveys' growth measurements). Overlaps
``"desi"``, as its BAO-only sibling does.
References
----------
eBOSS Collaboration / Alam et al. (2021), Phys. Rev. D 103,
083533, `arXiv:2007.08991 <https://arxiv.org/abs/2007.08991>`_.
"""
[docs]
def __init__(
self,
cosmology,
version="dr16",
):
dataset = load_sdss_fsbao(
version,
)
super().__init__(
name="SDSS-BAO+FS",
dataset=dataset,
cosmology=cosmology,
)