CosmoMC: help with DES likelihood and priors

 Posts: 22
 Joined: July 17 2014
 Affiliation: Johns Hopkins University
CosmoMC: help with DES likelihood and priors
I was looking at the DES files (batch3/DES.ini, data/DES/*) in the master branch and am having some trouble figuring out whether/how I can reproduce the DES or DES+Planck lensing results shown in the Planck 2018 papers. For example, what do I need to set as priors to reproduce the green, grey, and red contours from Fig 19 of Planck 2018 VI. https://arxiv.org/abs/1807.06209? In footnote 22 on page 27 it says "Here we assume consistent (DES) priors for DES and CMB lensing results". What are these priors? DES did not fit using the CosmoMC params (ombh2, omch2, theta etc.), right?
Sorry if there's some documentation on this and I missed it...
Sorry if there's some documentation on this and I missed it...

 Posts: 1435
 Joined: September 23 2004
 Affiliation: University of Sussex
 Contact:
Re: CosmoMC: help with DES likelihood and priors
You want to use
This switches the CosmoMC parameterization to the "astro" parameterization which can mimic what DES did.
Code: Select all
batch3/DES_astro_priors.ini

 Posts: 22
 Joined: July 17 2014
 Affiliation: Johns Hopkins University
Re: CosmoMC: help with DES likelihood and priors
Ah, thank you, I did indeed miss that.
Is there anywhere I could find a .covmat to use as a proposal matrix for the astro parameterization with DES? If not, and it's not too much trouble, could you send me one?
Is there anywhere I could find a .covmat to use as a proposal matrix for the astro parameterization with DES? If not, and it's not too much trouble, could you send me one?

 Posts: 22
 Joined: July 17 2014
 Affiliation: Johns Hopkins University
Re: CosmoMC: help with DES likelihood and priors
Two more questions (posting here instead of replying to your email, Antony, in case this is useful for other people reading the forum):
1) Does the DES likelihood include the corrections to the covariance matrix described in Appendix C of the DES https://arxiv.org/abs/1708.01530? I guess yes, since just from .likestats for the full 3x2pt shear+galaxygalaxy+galaxy clustering I get a chisquared around 500, and in the paper they say updating the covariance takes them from 572 to 497.
2) What's the best way to make a plot showing the DES data points and one or more theory curves (e.g. show two curves with two different theory models)? I see there's a function that seems to write the theory prediction called WL_WriteLikelihoodData in source/wl.f90. Do I just need to set test_output_root and run action=4?
1) Does the DES likelihood include the corrections to the covariance matrix described in Appendix C of the DES https://arxiv.org/abs/1708.01530? I guess yes, since just from .likestats for the full 3x2pt shear+galaxygalaxy+galaxy clustering I get a chisquared around 500, and in the paper they say updating the covariance takes them from 572 to 497.
2) What's the best way to make a plot showing the DES data points and one or more theory curves (e.g. show two curves with two different theory models)? I see there's a function that seems to write the theory prediction called WL_WriteLikelihoodData in source/wl.f90. Do I just need to set test_output_root and run action=4?

 Posts: 1435
 Joined: September 23 2004
 Affiliation: University of Sussex
 Contact:
Re: CosmoMC: help with DES likelihood and priors
1. Yes, it should have their final published covariance
2. I think you can do that. You can also use the python version under python/planck/DES.py with something like
(the full script I used to make the figure in the Planck parameter paper should be added to cosmomc when the +DES chains are finally published with the likelihoods)
2. I think you can do that. You can also use the python version under python/planck/DES.py with something like
Code: Select all
from planck.DES import DES_like
DESlens = DES_like(DESdataset_file, dataset_params={'used_data_types': 'xip xim'})
lens_theory = DESlens.get_theory_for_params(param_dictionary)
fig, axs = plt.subplots(DESlens.nzbins, DESlens.nzbins, figsize=(figsize, figsize), sharex='col', sharey='row')
for ax in axs.reshape(1): ax.axis('off')
for f1, f2 in DESlens.bin_pairs[0]:
ax = axs[f2, f1]
ax.axis('on')
xip = DESlens.data_arrays[0][f1, f2]
fac = 1e4 * DESlens.theta_bins
xip = xip * fac
ax.errorbar(DESlens.theta_bins, xip, fac * DESlens.errors[0][f1, f2], color='C0', fmt='.')
ax.semilogx(DESlens.theta_bins, fac * lens_theory[0][f1, f2], color='k', ls='')

 Posts: 22
 Joined: July 17 2014
 Affiliation: Johns Hopkins University
Re: CosmoMC: help with DES likelihood and priors
Oh, neat, I haven't been paying attention to the CosmoMC python stuff. Thanks! For the param_dictionary in this line is there a quick way to read a set of the DES nuisance params from DES chain files files (e.g. a .likestats or .minimum file)?
Code: Select all
lens_theory = DESlens.get_theory_for_params(param_dictionary)

 Posts: 1435
 Joined: September 23 2004
 Affiliation: University of Sussex
 Contact:
Re: CosmoMC: help with DES likelihood and priors
You can use e.g.
Code: Select all
from getdist.types import BestFit
param_dic=BestFit(r'chain_name.minimum', want_fixed=True).getParamDict()