data.over_sampling for PR4

[ming-jian]

Dear everyone,
I want to use the PR4 and other datasets on the montepython. However, it always wrong for the data.over_sampling.

data.experiments=['Planck20_Hillipop_TTTEEE', 'Planck20_Lollipop_EE', 'Planck_lowl_TT', 'BK18', 'Pantheon_Plus']

data.over_sampling=[1, 4, 2, 5]

I have used other data.over_sampling. The error is :

Configuration Error:
/|\ The length of the over_sampling field should be equal to the number of
/o\ blocks (one for cosmological parameters, plus one for each likelihood
with nuisance parameters)

Could you give me the correct data.over_sampling ? I also want to use the other datasets in the calculation. For example,

data.experiments=['Planck20_Hillipop_TTTEEE', 'Planck20_Lollipop_EE', 'Planck_lowl_TT', 'BK18', 'bao_desi_2025', 'Pantheon_Plus'].

[brinckmann]

Hi Ming-Jian,

You need to look at which likelihoods have nuisance parameters and which likelihoods share nuisance parameters. So e.g. if Planck20_Hillipop_TTTEEE and Pantheon_Plus have nuisance parameters but don't share them you need two additional entries beyond the cosmologyical parameters, i.e.:
data.over_sampling=[1, x, y]
So check your likelihoods and understand which ones have nuisance parameters.

Additionally, you should put likelihoods that share nuisance parameters next to each other in the likelihood list, so MontePython understands to group them, so e.g. if any of the likelihoods share a nuisance parameter with Planck Hillipop you need to put them next in the list, but you should only have one entry in over_sampling for both of them.

Let me know if these instructions don't help and I can try to have a look at the specific likelihoods.

Best,
Thejs

[ming-jian]

Dear Prof. Thejs,
Thanks very much for your kind help! The param is shown here:

data.experiments=['Planck20_Hillipop_TTTEEE','Planck_lowl_TT','Planck20_Lollipop_EE', 'BK18', 'Pantheon_Plus']

#------ Settings for the over-sampling.

The first element will always be set to 1, for it is the sampling of the

cosmological parameters. The other numbers describe the over sampling of the

nuisance parameter space. This array must have the same dimension as the

number of blocks in your run (so, 1 for cosmological parameters, and then 1

for each experiment with varying nuisance parameters).

Note that when using Planck likelihoods, you definitely want to use [1, 4],

to oversample as much as possible the 14 nuisance parameters.

Remember to order manually the experiments from slowest to fastest (putting

Planck as the first set of experiments should be a safe bet, except if you

also have LSS experiments).

If you have experiments without nuisance, you do not need to specify an

additional entry in the over_sampling list (notice for instance that, out of

the three Planck likelihoods used, only Planck_highl requires nuisance

parameters, therefore over_sampling has a length of two (cosmology, plus one

set of nuisance).

data.over_sampling=[1, 4, 2, 5]

#------ Parameter list -------

data.parameters[class name] = [mean, min, max, 1-sigma, scale, role]

- if min max irrelevant, put to -1 or None (if you want a boundary of -1, use -1.0)

- if fixed, put 1-sigma to 0

- if scale irrelevant, put to 1, otherwise to the appropriate factor

- role is either 'cosmo', 'nuisance' or 'derived'

Cosmological parameters list

data.parameters['omega_b'] = [0.0223, 0.017, 0.027, 0.0002, 1, 'cosmo']
data.parameters['omega_cdm'] = [0.1188, 0.09, 0.15, 0.002, 1, 'cosmo']
data.parameters['n_s'] = [ 0.968, 0.9, 1.1, 0.005, 1, 'cosmo']
data.parameters['ln10^{10}A_s'] = [ 3.040, 2.7, 3.4, 0.02, 1, 'cosmo']
data.parameters['tau_reio'] = [ 0.058, 0.01, 0.17, 0.008, 1, 'cosmo']
data.parameters['100*theta_s'] = [1.0410, 1.03, 1.05, 0.0003, 1, 'cosmo']

Nuisance parameter list, same call, except the name does not have to be a class name

Note the nuisance parameters must follow immediately after the cosmo parameters,

MP expects cosmo, nuisance, derived

Nuisance parameters

#calib_params
data.parameters['A_planck'] = [ 1.0, 0.9, 1.1, 0.0025, 1, 'nuisance']
data.parameters['cal100A'] = [ 1.0, 0.9, 1.1, 0.005, 1, 'nuisance']
data.parameters['cal100B'] = [ 1.0, 0.9, 1.1, 0.005, 1, 'nuisance']
data.parameters['cal143A'] = [ 1.0, 1, 1, 0, 1, 'nuisance']
data.parameters['cal143B'] = [ 1.0, 0.9, 1.1, 0.005, 1, 'nuisance']
data.parameters['cal217A'] = [ 1.0, 0.9, 1.1, 0.005, 1, 'nuisance']
data.parameters['cal217B'] = [ 1.0, 0.9, 1.1, 0.005, 1, 'nuisance']

#TT
data.parameters['Aradio'] = [ 63.3, 0, 150, 4.7, 1, 'nuisance']
data.parameters['Adusty'] = [ 6.07, 0, 100, 0.6, 1, 'nuisance']
data.parameters['AdustT'] = [ 1.0, 0.5, 1.5, 0.03, 1, 'nuisance']
data.parameters['beta_dustT'] = [ 1.51, 1.2, 1.8, 0.01, 1, 'nuisance']
data.parameters['Acib'] = [ 1.0, 0., 20, 0.3, 1, 'nuisance']
data.parameters['Atsz'] = [ 5.9, 0., 50, 1.6, 1, 'nuisance']
data.parameters['Aksz'] = [ 1.0, 0., 50, 2.5, 1, 'nuisance']
data.parameters['xi'] = [ 0.4, -1., 1., 0.3, 1, 'nuisance']
data.parameters['beta_cib'] = [ 1.75, 1.6, 1.9, 0.06, 1, 'nuisance']
data.parameters['beta_radio'] = [ -0.8, -0.8, -0.8, 0, 1, 'nuisance']
data.parameters['beta_dusty'] = [ 1.75, 1.6, 1.9, 0.06, 1, 'nuisance']

#EE
data.parameters['AdustP'] = [ 1.0, 0.5, 1.5, 0.03, 1, 'nuisance']
data.parameters['beta_dustP'] = [ 1.58, 1.2, 1.8, 0.02, 1, 'nuisance']

#TE
#data.parameters['AdustT'] = [ 1.0, 0.5, 1.5, 0.03, 1, 'nuisance']
#data.parameters['beta_dustT'] = [ 1.51, 1.2, 1.8, 0.01, 1, 'nuisance']
#data.parameters['AdustP'] = [ 1.0, 0.5, 1.5, 0.03, 1, 'nuisance']
#data.parameters['beta_dustP'] = [ 1.58, 1.2, 1.8, 0.02, 1, 'nuisance']

dust power at ell=80, nu=353 GHz [uK^2]

data.parameters['BBdust'] = [ 3., 0., 15., 0.1, 1, 'nuisance']

sync power at ell=80, nu=23 GHz [uK^2]

data.parameters['BBsync'] = [ 1., 0., 50., 1., 1, 'nuisance']

dust spatial power spectrum power law index

data.parameters['BBalphadust'] = [ -0.42, -1.0, 0., 0.01, 1, 'nuisance']

dust SED power law index, for use with Planck relax prior and set bound [0.80, 2.40]

#data.parameters['BBbetadust'] = [ 1.59, 1.04, 2.14, 0.02, 1, 'nuisance']
data.parameters['BBbetadust'] = [ 1.59, 0.8, 2.4, 0.02, 1, 'nuisance']

dust blackbody temperature [K] -- fixed / very insensitive to this

data.parameters['BBTdust'] = [ 19.6, 19.6, 19.6, 0.0, 1, 'nuisance']

sync spatial power specturm power law index

data.parameters['BBalphasync'] = [ -0.6, -1.0, 0., 0.01, 1, 'nuisance']

sync SED power law index

data.parameters['BBbetasync'] = [ -3.1, -4.5, -2.0, 0.02, 1, 'nuisance']

correlation between dust and sync

NOTE: For BK15 analysis, we are allowing this parameter to range

from -1 to 1 (allowed range was 0 to 1 for BK14).

data.parameters['BBdustsynccorr'] = [ 0.2, -1.0, 1.0, 0.01, 1, 'nuisance']

dust/sync correlation ratio between 217 and 353 GHz, ell=80, fixed to 1.0 = no decorrelation

data.parameters['Delta_dust'] = [ 1.0, 1.0, 1.0, 0, 1, 'nuisance']
data.parameters['Delta_sync'] = [ 1.0, 1.0, 1.0, 0, 1, 'nuisance']

EE/BB ratios -- fixed / only used if E-modes are turned on

data.parameters['EEtoBB_dust'] = [ 2.0, 2.0, 2.0, 0, 1, 'nuisance']
data.parameters['EEtoBB_sync'] = [ 2.0, 2.0, 2.0, 0, 1, 'nuisance']

Band center errors, fixed to zero

data.parameters['gamma_corr'] = [ 0.0, 0.0, 0.0, 0, 1, 'nuisance']
data.parameters['gamma_95'] = [ 0.0, 0.0, 0.0, 0, 1, 'nuisance']
data.parameters['gamma_150'] = [ 0.0, 0.0, 0.0, 0, 1, 'nuisance']
data.parameters['gamma_220'] = [ 0.0, 0.0, 0.0, 0, 1, 'nuisance']

data.parameters['M'] = [-19.2, -30, -10, 0.5, 1, 'nuisance']

Derived parameter list

data.parameters['z_reio'] = [0, None, None, 0, 1, 'derived']
data.parameters['Omega_Lambda'] = [0, None, None, 0, 1, 'derived']
data.parameters['A_s'] = [0, None, None, 0, 1e-9, 'derived']
data.parameters['H0'] = [0, None, None, 0, 1, 'derived']
#data.parameters['sigma8'] = [0, None, None, 0, 1, 'derived']

CLASS parameters

data.cosmo_arguments['sBBN file'] = data.path['cosmo']+'/external/bbn/sBBN_2017.dat'
data.cosmo_arguments['k_pivot'] = 0.05
data.cosmo_arguments['N_ur'] = 2.0328
data.cosmo_arguments['N_ncdm'] = 1
data.cosmo_arguments['m_ncdm'] = 0.06
data.cosmo_arguments['T_ncdm'] = 0.71611
data.cosmo_arguments['l_max_scalars'] = 2500
data.cosmo_arguments['non_linear'] = 'halofit'

These two are required to get sigma8 as a derived parameter

(class must compute the P(k) until sufficient k)

#data.cosmo_arguments['output'] = 'mPk'
#data.cosmo_arguments['P_k_max_h/Mpc'] = 1.

#------ Mcmc parameters ----

Number of steps taken, by default (overwritten by the -N command)

data.N=10

Number of accepted steps before writing to file the chain. Larger means less

access to disc, but this is not so much time consuming.

data.write_step=5

Note: MP will only interpret comments if the line begins with # (you cannot add comments at the end of lines!)

[brinckmann]

I agree you need four entries in over_sampling (1 for cosmo 3 for nuisance). As far as I can tell you have four likelihoods with nuisance parameters, Planck20_Hillipop_TTTEEE, Planck20_Lollipop_EE, BK18 and Pantheon_Plus. As far as I can see from the likelihood's .data file the Planck20_Lollipop_EE uses A_planck, which it shares with Planck20_Hillipop_TTTEEE, so I think it should be next to the Hillipop likelihood, since I don't think lowl_TT uses a nuisance parameter.

I can't remember if A_planck also needs to be at the end of the list of Planck nuisance parameters in the param file in order to indicate it's shared with the next likelihood (I don't think so, but it wouldn't hurt).

Also, just in case, but make sure all the BK18 nuisance parameters are after the Planck nuisance parameters in the param file. Does BK18 share any nuisance parameters with Planck or are they all separate? I've never used the BK likelihood, so I'm not familiar with it.

The rule of thumb is to group your nuisance parameters in your param file in the same order that the likelihoods are listed in the likelihood list, with likelihoods that share nuisance parameters next to each other and the shared nuisance parameters between their individual lists of nuisance parameters. I'm not sure if the last point matters, but I always do it.

Best,
Thejs