New all_l_exact data format?

[Will Kinney]

I'm trying to add some synthetic data in all_l_exact format, using the procedure outlined in this old thread:

http://cosmocoffee.info/viewtopic.php?t=231

I'm getting errors from CMBlikes.f90 such as:

Error for key "fields_use" : key not found

I notice in the header there is a comment:

!Mar 2014 renaned from Planck_like, removed low-L likelihood to separate file
!Updated file structure, allows for binned HL likelihoods. Everything now in L(L+1)CL/2pi

Presumably there is a new format for .dataset files? Is there documentation of the new file format any place? (Also, getting "renaned" sounds vaguely unpleasant ;-)

Thanks for any pointers.

[Antony Lewis]

Try adapting this python script (which in the next update will replace python/makePerfectForecastDataset.py; note you need the python directory in your path to find e.g.iniFile.py)

Code: Select all

# take CAMB file (e.g. test_lensedCls.dat) and produce dataset with given noise for testing
# Use in cosmomc .ini file using e.g.
# cmb_dataset[MyForecast]=data/MyForecast/test_lensedCls_exactsim.dataset

from numpy import *
import iniFile
import shutil
import sys, os


# Edit parameters you want to change here

root = r'test_lensedCls'

lensedTotClFileRoot = r'c:\work\dist\git\camb\test_lensedCls'
outDir = 'z:\\data\\'

fwhm_arcmin = 7.
# Noise var is N_l in muK^2 for white noise
NoiseVar = 2e-4
# Pol noise var = ENoiseFac * NoiseVar
ENoiseFac = 4
lmin = 2
lmax = 2000
fsky = 1

DoLensing = False
# if including an idealised lensing reconstruction
reconNoise = 'reconNoise.dat'
lensPotential = 'flatLCDM_lenspotentialCls.dat'


# os.path.dirname(sys.path[0])+'/data/'
ini = iniFile.iniFile()
dataset = ini.params

# change this if you don't want to use all pol
dataset['fields_use'] = 'T E'


# #Now produce the Planck_like files
fwhm = fwhm_arcmin / 60
xlc = 180 * sqrt(8.*log(2.)) / pi
sigma2 = (fwhm / xlc) ** 2

outPath = ''
outRoot = root + '_exactsim'
NoiseOut = []
for l in range(lmax + 1):
    NoiseCl = l * (l + 1) / 2 / pi * NoiseVar * exp(l * (l + 1) * sigma2)
    NoiseOut.append([NoiseCl, ENoiseFac * NoiseCl, ENoiseFac * NoiseCl])


outfile = open(outDir + outRoot + '_Noise.dat', 'w')
for l in range(2, lmax + 1):
    outfile.write("%d " % l + " ".join("%E" % elem for elem in NoiseOut[l]) + "\n")
outfile.close()

dataset['fullsky_exact_fksy'] = fsky
dataset['name'] = outRoot
dataset['dataset_format'] = 'CMBLike2'
dataset['like_approx'] = 'exact'

dataset['cl_lmin'] = lmin
dataset['cl_lmax'] = lmax

dataset['binned'] = False


dataset['cl_hat_includes_noise'] = False

dataset['cl_hat_file'] = outPath + outRoot + '.dat'
dataset['cl_hat_order'] = 'TT EE BB TE'
dataset['cl_noise_file '] = outPath + outRoot + '_Noise.dat'
dataset['cl_noise_order'] = 'TT EE BB'

if DoLensing:
    dataset['lensing_recon_ncl'] = 1
    dataset['cl_hat_phi_file'] = outPath + outRoot + '_PhiEst.dat'
    dataset['cl_noise_phi_file'] = outPath + outRoot + '_PhiNoise.dat'
    dataset['cl_hat_includes_noise'] = False
    shutil.copy(reconNoise, outDir + outRoot + '_PhiNoise.dat')
    fout = open(outDir + outRoot + '_PhiEst.dat', 'w')
    f = open(lensPotential)
    for line in f:
            row = line.split()
            fout.write(str(int(row[0])) + ' ' + str(row[5]) + '\n')
    f.close()
    fout.close()


ini.saveFile(outDir + outRoot + '.dataset')

shutil.copy(lensedTotClFileRoot + '.dat', outDir + outRoot + '.dat')

Note I haven't tested recently if lensing reconstruction exact likelihood works.

[Will Kinney]

Thanks, Antony. This works, except I had to add a 'fields_use' field to avoid an error, as in:

Code: Select all

dataset['fields_use'] = 'TEB'

Sound right?

[Antony Lewis]

The example already has

Code: Select all

dataset['fields_use'] = 'T E' 

?