Merge pull request #99 from joezuntz/kids-only-like

Add kids-only cosebis likelihood

examples/kids-1000-values.ini

@@ -0,0 +1,38 @@
+;In KiDS-1000 the photo-z uncertainty is assumed to be correlated across the z-bins (see prior file)
+[nofz_shifts_kids]
+uncorr_bias_1 = -5.0   0.000  5.0
+uncorr_bias_2 = -5.0  -0.181  5.0
+uncorr_bias_3 = -5.0  -1.110  5.0
+uncorr_bias_4 = -5.0  -1.395  5.0
+uncorr_bias_5 = -5.0   1.265  5.0
+
+; The Hybrid pipeline set up uses independent IA parameters for each survey
+; Here we use Tophat priors for an NLA-z analysis, but if you wanted to analyse
+; TATT you can modify the priors on the A2, alpha2 and bias_ta parameters
+[intrinsic_alignment_parameters]
+z_piv   =  0.62
+A1      = -5.0   0.0  5.0
+A2      = 0.0
+alpha1  = -5.0  0.0  5.0
+alpha2  = 0.0
+bias_ta =  0.0
+
+
+; We use a Tophat prior to marginalise over our uncertainty on baryon feedback using HMCode2020
+[halo_model_parameters]
+logt_agn = 7.3 7.8 8.0
+
+; This is the set of cosmological parameter priors that were found to 
+; introduce the least projection effects on the marginal S8 distribution
+; Note that CosmoSIS v3 onwards can sample over S8
+[cosmological_parameters]
+omch2 = 0.051    0.11812972217650827    0.255
+ombh2 = 0.019    0.025939374402978773    0.026
+h0 =    0.64     0.7666550530735352    0.82
+n_s =  0.84    0.9007697522848085    1.1
+S_8 = 0.1  0.7567464875805479    1.3
+omega_k = 0.0
+w       = -1.0
+wa      = 0.0
+mnu     = 0.055  0.07740741  0.6
+tau     =  0.0697186

examples/kids-1000.ini

@@ -0,0 +1,175 @@
+; We will carry out a joint Hybrid pipeline analysis of 
+; DES Y3 xi_pm and KiDS-1000 COSEBIS including scale cuts (2'<theta_min<300')
+[DEFAULT]
+DATAFILE=likelihood/des-y3_and_kids-1000/DES-Y3_xipm_and_KiDS-1000_COSEBIs_2.0_300.0.fits
+RUN_NAME = des-y3_and_kids-1000
+
+[pipeline]
+modules =   consistency camb extrapolate 
+             correlated_dz_priors fits_nz_kids photoz_bias_kids 
+            fast_pt IA pk_to_cl_kids add_intrinsic cosebis cosebis_like
+likelihoods = cosebis
+values  = examples/kids-1000-values.ini
+priors  = examples/des-y3_and_kids-1000-priors.ini
+extra_output =  cosmological_parameters/sigma_8 
+                cosmological_parameters/A_s cosmological_parameters/omega_m 
+                cosmological_parameters/omega_lambda cosmological_parameters/cosmomc_theta 
+                delta_z_out_kids/bin_1 delta_z_out_kids/bin_2 delta_z_out_kids/bin_3 
+                delta_z_out_kids/bin_4 delta_z_out_kids/bin_5  
+                likelihoods/cosebis_like
+timing = F
+debug = T
+
+; Since CosmoSIS v3, the consistency interface allows for sampling over S8
+; Here we set up the non-linear power spectrum
+[consistency]
+file = utility/consistency/consistency_interface.py
+cosmomc_theta=T
+
+[camb]
+file = boltzmann/camb/camb_interface.py
+mode = all
+halofit_version = mead2020_feedback  
+neutrino_hierarchy = normal
+lmax=2500
+kmax=100.0
+zmid = 2.0
+nz_mid = 100
+zmax = 6.0
+nz = 150
+feedback=0
+
+[extrapolate]
+file = boltzmann/extrapolate/extrapolate_power.py
+kmax = 500.
+
+; Next we define the redshift bins and how we're going to marginalise over
+; our uncertainty on these distributions
+[fits_nz_des]
+file = number_density/load_nz_fits/load_nz_fits.py
+nz_file = %(DATAFILE)s
+data_sets = source_des
+
+; This module allows for correlated priors for KiDS given a covariance matrix
+[correlated_dz_priors]
+file = number_density/correlated_priors/correlated_priors.py
+uncorrelated_parameters =   nofz_shifts_kids/uncorr_bias_1 nofz_shifts_kids/uncorr_bias_2 
+                            nofz_shifts_kids/uncorr_bias_3 nofz_shifts_kids/uncorr_bias_4 
+                            nofz_shifts_kids/uncorr_bias_5
+output_parameters = nofz_shifts_kids/bias_1 nofz_shifts_kids/bias_2 
+                    nofz_shifts_kids/bias_3 nofz_shifts_kids/bias_4 
+                    nofz_shifts_kids/bias_5
+covariance = likelihood/des-y3_and_kids-1000/nofz_covariance/SOM_cov_multiplied.asc
+
+[fits_nz_kids]
+file = number_density/load_nz_fits/load_nz_fits.py
+nz_file = %(DATAFILE)s
+data_sets = source_kids
+
+
+[photoz_bias_kids]
+file = number_density/photoz_bias/photoz_bias.py
+mode = additive
+sample = nz_source_kids
+bias_section = nofz_shifts_kids
+interpolation = cubic
+output_deltaz_section_name = delta_z_out_kids
+
+; Here we are using the TATT modules for our IA model
+; to allow for flexibility in extensions to our fiducial analyses
+; The hybrid set-up uses NLA-z, with the non-NLA parameters
+; in the TATT model set to zero in the values file
+
+[fast_pt]
+file = structure/fast_pt/fast_pt_interface.py
+do_ia = T
+k_res_fac = 0.5
+verbose = F
+
+
+[IA]
+file = intrinsic_alignments/tatt/tatt_interface.py
+sub_lowk=F
+do_galaxy_intrinsic=F
+ia_model=tatt
+
+; As DES and KiDS are assumed to be uncorrelated we can 
+; calculate the likelihoods independently
+
+
+[add_intrinsic]
+file=shear/add_intrinsic/add_intrinsic.py
+shear-shear=T
+position-shear=F
+perbin=F
+
+
+[pk_to_cl_kids]
+file = structure/projection/project_2d.py
+ell_min_logspaced = 0.1
+ell_max_logspaced = 5.0e5
+n_ell_logspaced = 100 
+shear-shear = source_kids-source_kids
+shear-intrinsic = source_kids-source_kids
+intrinsic-intrinsic = source_kids-source_kids
+intrinsicb-intrinsicb = source_kids-source_kids
+verbose = F
+get_kernel_peaks = F
+sig_over_dchi = 20. 
+shear_kernel_dchi = 10.
+
+;This calculates COSEBIs from Cls
+[cosebis]
+file=shear/cosebis/cl_to_cosebis/cl_to_cosebis_interface.so
+theta_min = 2.0 ; default=0.5
+theta_max = 300.0 ; default=300
+n_max = 5 ; default=5
+input_section_name = shear_cl
+output_section_name = cosebis
+
+;This is a simplified version of the the KiDS-1000 2pt_likelihood module
+;which is not all-singing nor all-dancing because we only need to
+;analyse COSEBIs for this Joint DES+KiDS analysis
+[cosebis_like]
+file = likelihood/2pt/cosebis/simple_like.py
+data_set = En n
+data_file = %(DATAFILE)s
+like_name = cosebis
+;If you're interested in using KiDS-1000 data products other than COSEBIs
+;The 2pt-likelihood that KiDS-1000 used for their 3x2pt analysis can be found
+;on the KiDS-1000 KCAP repo: 
+;https://github.com/KiDS-WL/kcap/blob/master/modules/scale_cuts/scale_cuts.py
+;https://github.com/KiDS-WL/kcap/blob/master/utils/mini_like.py
+
+
+; If I run on the command line "cosmosis examples/des-y3_and_kids-1000.ini", this will run this quick test sampler.
+[runtime]
+sampler = test
+verbosity = standard
+; saving the output to output/des-y3_and_kids-1000
+[test]
+save_dir=output/%(RUN_NAME)s
+fatal_errors=T
+
+; If I run command "cosmosis examples/des-y3_and_kids-1000.ini runtime.sampler='polychord'", this will run polychord.
+; These are the settings used for the Hybrid DES+KiDS analysis, but be warned it's CPU intensive. 
+[polychord]
+base_dir = chain_checkpoints
+polychord_outfile_root=poly_%(RUN_NAME)s
+resume=F
+feedback = 3
+fast_fraction = 0.1
+;Minimum settings for a "good enough" quick test
+;live_points = 250
+;num_repeats = 30
+;tolerance = 0.1
+;Settings for high quality paper runs
+live_points = 500
+num_repeats=60
+tolerance=0.01
+boost_posteriors=10.0
+
+[output]
+filename= output/kids-1000.txt
+format=text
+privacy=F

tests/test_cosmosis_standard_library.py

@@ -99,3 +99,8 @@ def test_log_w_example():
 def test_des_kids(capsys):
     run_cosmosis("examples/des-y3_and_kids-1000.ini")
     check_likelihood(capsys, "-199.40", "-199.41")
+
+
+def test_kids(capsys):
+    run_cosmosis("examples/kids-1000.ini")
+    check_likelihood(capsys, "-47.6")