sacc_like

Generic 2-point measurement Gaussian likelihood using sacc format

File

likelihood/sacc/sacc_like.py

Attribution

CosmoSIS Team

URL

This module implements a general likelihood of tomographic 2-point measuremnts of various quantities, including galaxy density, cosmic shear, intrinsic alignments, CMB lensing, and the various cross powers between these.

This is a first version of a general code using the SACC format developed for DESC. See github.com/LSSTDESC/sacc for more details

Note a recent change: by default we now do not renormalize any bandpower weights, whereas before we did. This is to allow for non-unit-sum weights such as those produced by NaMaster.

Assumptions

  • A Gaussian likelihood approximation for two-point measurements

  • Data supplied in a specific file format

Setup Parameters

Name

Type

Default

Description

data_file

str

Filename of the sacc file to use.

data_sets

str

all

Space-separated list of which data sets from within the file to use for likelihoods.

keep_tracers

str

Regular expression to select tracers to use.

angle_range_{dataset}_{i}_{j}

str

Pair of real numbers. If set, for the given data set and pair of bins, cut down the data used to this angular range (min and max)

cut_{dataset}

str

Space-separated list of i,j pairs. (no spaces within the pair, just betwen them, e.g. cut_lss = 1,2 1,1 3,4. Remove this bin from the likelihood.

{name}_section

str

Various depending on name.

For each {name} in the data types used from the file, a cosmosis block section to look for the theory predictions in.

save_theory

str

If set, save the theory predictions used in the likelihood to this sacc file.

save_realization

str

If set, save a simulated data set to this sacc file.

covariance_realizations

int

-1

If >0, assume that the Covariance matrix was estimated from a set of MC simulations and should thus have the Anderson-Hartlap factor applied to increase its size. If zero, assume infinite number of realizations.

sellentin

bool

False

If set, use the Sellentin-Heavens 2016 change to the likelihood to account for this distribution of the covariance estimates. This changes the likelihood to a student’s-t form. Note that this invalidates the simulated data sets used for the ABC sampler.

like_name

str

2pt

The name of the likelihood to save.

likelihood_only

bool

False

Skip saving the covariance, inverse, simulation, etc. Saves some time.

kind

str

cubic

The interpolation to do into the theory splines. See scipy.interpolate.interp1d.

Input values

Section

Name

Type

Default

Description

shear_cl

ell

real 1d

If a Fourier-space measurement is used, the angular wave-number of the predicted theory curves. The name of the section here depends on the data type used from the file. It might be galaxy_cl or shear_cl, for example.

bin_{i}_{j}

real 1d

For various i,j depending what is found in the file, the theory predictions for this value. For example, C_ell.

Output values

Output values

Section

Name

Type

Description

likelihoods

2pt_like

real

Gaussian likelihood value. Name can be changed in parameter file (see above) for this and the other outputs below.

data_vector

2pt_data

real 1d

The full vector of data points used in the likelihood

2pt_theory

real 1d

The full vector of theory points used in the likelihood

2pt_covariance

real 2d

The covariance matrix used

2pt_inverse_covariance

real 2d

The inverse covariance matrix (precision matrix) used.

2pt_simulation

real 1d

A simulated data set from the given theory and covariance matrix.

2pt_angle

real 1d

The angular scale used for each data point.

2pt_bin1

int 1d

The first bin index used for each data point

2pt_bin2

int 1d

The second bin index used for each data point