Data Sets

Data Sets#

Data sets for candl are kept separately from the code. There currently exist three library repositories with compatible data and you can find detailed information on them below. You can still use candl without any data library for the different tools supplied or to create your own likelihoods.

Warning

For versions v1.* of candl data came directly with the pip installation. This is no longer the case (due to the growing size of amazing CMB data we have!) and the data need to be installed separately of the code.

spt_candl_data#

Official repository of the South Pole Telescope collaboration featuring the latest SPT-3G data. Simply clone the repo to get the data and run pip install . inside the folder to install. You can use spt_candl_data.print_all_shortcuts to list all available shortcuts to data sets. More details can be found in the spt_candl_data readme.

SPT-3G D1 T&E#

Paper(s):
E. Camphuis, W. Quan, L. Balkenhol, A. R. Khalife, F. Ge, F. Guidi, N. Huang, G. P. Lynch, Y. Omori, C. Trendafilova et al. 2025 (SPT-3G Collaboration)
spt3gd1tne
W. Quan et al. (SPT-3G Collaboration), in prep.
Type:

Primary power spectrum measurement (\(TT/TE/EE\))

Website:

SPT Website

LAMBDA:

To come.

Short cut(s):
spt_candl_data.SPT3G_D1_TnE (index file),
spt_candl_data.SPT3G_D1_TnE_multifreq (or variant = 'multifreq'),
spt_candl_data.SPT3G_D1_TnE_lite (or variant = 'lite')
Latest version:

v0

Tip

For this likelihood, please use dedicated variants if you are only fitting a subset of the data (e.g. only EE spectra or only 90GHz) and only use the candl functionality data_selection for multipole cuts. See all available variants via spt_candl_data.print_all_shortcuts().

SPT-3G D1 BB#

Paper(s):
J. A. Zebrowski, C. L. Reichardt et al. 2025 (SPT-3G Collaboration)
spt3gd1bb
Type:

Primary power spectrum measurement (\(BB\))

Website:

SPT Website

LAMBDA:

To come.

Short cut(s):
spt_candl_data.SPT3G_D1_BB
Latest version:

v0

Tip

This likelihood uses the Hamimeche-Lewis approximation. As such, not all subsets of the data set can be run. You may also consider the foreground-marginalised (CMB-only) version of this data set in candl_data.

candl_data#

The data below are made available through the candl_data repository. Simply clone the repo to get the data and run pip install . inside the folder to install. You can use candl_data.print_all_shortcuts() to list all available shortcuts to data sets. More details can be found in the candl_data readme.

SPT-3G 2018 TT/TE/EE#

Paper(s):
L. Balkenhol, D. Dutcher, A. Spurio Mancini, A. Dussot, K. Benabed, S. Galli et al. (SPT-3G Collaboration)
spt3g18ttteee
Type:

Primary power spectrum measurement (\(TT/TE/EE\))

Website:

SPT Website

LAMBDA:

NASA Archive

Short cut(s):
candl_data.SPT3G_2018_TTTEEE (index file),
candl_data.SPT3G_2018_TTTEEE_multifreq (or variant = 'multifreq'),
candl_data.SPT3G_2018_TTTEEE_lite (or variant = 'lite')
Latest version:

v0

Tip

Though the multi-frequency likelihood is loaded by default, be sure to check out the much faster lite version with variant = 'lite' introduced in Balkenhol 2024. If you are interested in how the lite likelihood is constructed, see the associated notebook in the tutorials.

SPT-3G 2018 PP#

Paper(s):
Z. Pan, F. Bianchini, W. L. K. Wu et al. (SPT-3G Collaboration)
spt3g18pp
Type:

Lensing power spectrum measurement (\(\phi\phi\))

Short cut(s):
candl_data.SPT3G_2018_Lens (index file),
candl_data.SPT3G_2018_Lens_only (or variant = 'lens_only'),
candl_data.SPT3G_2018_Lens_and_CMB (or variant = 'use_CMB')
Latest version:

v0

Note: this data set uses the lensing power spectrum in \(\phi\phi\). Use candl_data.SPT3G_2018_Lens with variant = 'lens_only' or candl_data.SPT3G_2018_Lens_only when only working with lensing data, use candl_data.SPT3G_2018_Lens with variant = 'use_CMB' or candl_data.SPT3G_2018_Lens_and_CMB when combining lensing and primary CMB data.

SPT-3G D1 BB lite#

Paper(s):
J. A. Zebrowski, C. L. Reichardt et al. 2025 (SPT-3G Collaboration)
spt3gd1bb
L. Balkenhol, A. Coerver, C. L. Reichardt, and J. A. Zebrowski 2025
spt3gd1bblite
Type:

Primary power spectrum measurement (\(BB\))

Website:

SPT Website

Short cut(s):
candl_data.SPT3G_D1_BB_lite
Latest version:

v0

Note: Foreground-marginalised (CMB-only) version of the spt_candl_data.SPT3G_D1_BB likelihood. This data set is not suited for constraining primordial signals that do not match the CMB black body SED.

SPTpol BB#

Paper(s):
J. T. Sayre, C. L. Reichardt, J. W. Henninget al. 2020 (SPTpol Collaboration)
sptpolbb
L. Balkenhol, A. Coerver, C. L. Reichardt, and J. A. Zebrowski 2025
spt3gd1bblite
Type:

Primary power spectrum measurement (\(BB\))

Website:

SPT Website

LAMBDA:

NASA archive

Short cut(s):
candl_data.SPTpol_BB (index file),
candl_data.SPTpol_BB_multifreq (or variant = 'multifreq'),
candl_data.SPTpol_BB_lite (or variant = 'lite')
Latest version:

v0

Note: This is an implementation of the original Fortran likelihood. Please see Balkenhol et al. 2025 for details, especially when scrutinising \(\chi^2\) values. Note that the CMB-only version of this data set is not suited for constraining primordial signals that do not match the CMB black body SED.

ACT DR6 TT/TE/EE#

Paper(s):
Naess, Guan, Duivenvoorden, Hasselfield, Wang et al. (ACT Collaboration)
actdr6ttteee_naess
Louis, La Posta, Atkins, Jense et al. (ACT Collaboration)
actdr6ttteee_louis
Calabrese, Hill, Jense, La Posta et al. (ACT Collaboration)
actdr6ttteee_calabrese
Type:

Primary power spectrum measurement (\(TT/TE/EE\))

Website:

ACT Website ACT DR6 Notebooks

LAMBDA:

NASA archive

Short cut(s):

candl_data.ACT_DR6_TTTEEE

Latest version:

v0

Note: This is the CMB-only, foreground marginalised (lite) version of the likelihood. The ACT collaboration suggests to combine with the sroll2 likelihood to constrain the optical depth to reionisation; this implementation of the likelihood contains by default their suggested alternative \(\tau\) prior of \(0.0566 \pm 0.0058\). If you use ACT data via candl, please see the attribution instructions in the ACT DR6 notebooks README.

ACT DR4 TT/TE/EE#

Paper(s):
S. Aiola, E. Calabrese, L. Maurin, S. Naess, B. L. Schmitt et al. (ACT Collaboration)
actdr4ttteee_aiola
S. K. Choi, M. Hasselfield, S.-P. P. Ho, B. Koopman, M. Lungu et al. (ACT Collaboration)
actdr4ttteee_choi
Type:

Primary power spectrum measurement (\(TT/TE/EE\))

LAMBDA:

NASA archive

Short cut(s):

candl_data.ACT_DR4_TTTEEE

Latest version:

v0

Note: This is the CMB-only, foreground marginalised (lite) version of the likelihood. The likelihood refers to the deep data as dxd and the wide data as wxw.

ACT DR6 PP#

Paper(s):
M. S. Madhavacheril, F. J. Qu, B. D. Sherwin, N. MacCrann, Y. Li et al. (ACT Collaboration)
actdr4pp_madhavacheril
F. J. Qu, B. D. Sherwin, M. S. Madhavacheril, D. Han, K. T. Crowley et al. (ACT Collaboration)
actdr4pp_qu
Type:

Lensing power spectrum measurement (\(\phi\phi\))

Website:

Github

Short cut(s):
candl_data.ACT_DR6_Lens (index file),
candl_data.ACT_DR6_Lens_only (or variant = 'lens_only'),
candl_data.ACT_DR6_Lens_and_CMB (or variant = 'use_CMB')
Latest version:

v1

Note: this data set uses the lensing power spectrum in \(\kappa\kappa\). For the ACT + Planck lensing combination see also Carron, Mirmelstein, Lewis 2023. Use candl_data.ACT_DR6_Lens with variant = 'lens_only' or candl_data.ACT_DR6_Lens_only when only working with lensing data, use candl_data.ACT_DR6_Lens with variant = 'use_CMB' or candl_data.ACT_DR6_Lens_and_CMB when combining lensing and primary CMB data.

Planck#

candl comes with clipy, a pure Python implementation of the 2018 Planck likelihoods. See the clipy website to see which specific likelihoods supported and be sure to download the respective data files from the Planck Legacy Archive.

The Planck likelihoods are not implemented as native candl likelihoods, but as wrappers. While they are differentiable and work with a lot of the candl tools and interface code, they do not support the full functionality. See the clipy x candl tutorial for a demonstration of what’s possible.

Adding Your Own Data Sets#

If you wish to install data sets separately from the code, please download the desired folders individually from the GitHub repo. You can place these wherever you like. It simply suffices to point to the .yaml file of a likelihood to initialise it. If you wish to build your own data sets, please consult the information here.