Configuring the Theory¶
Even if the user chooses to use the default theory parameters returned
by GalaxySpectrum.default_params(), there are several areas where the user
should customize the parameters for the specific data set being fit. We will
described these customizations in this section.
Changing the Parametrization¶
The recommended way to make changes to the model parametrization is by
adjusting the parameters in the default
pyRSD.rsdfit.theory.parameters.ParameterSet object. The user
should get the default parameters from the default_params() function
and then make the desired changes, before writing the parameters to a file.
For example, to fix the satellite fraction and only use 12 free parameters in the fitting procedure, one can simply do
In [1]: from pyRSD.rsd import GalaxySpectrum
In [2]: model = GalaxySpectrum()
# default params
In [3]: params = model.default_params()
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-3-20168f551e3a> in <module>()
----> 1 params = model.default_params()
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsd/power/gal/power_gal.py in default_params(self)
128 """
129 from pyRSD.rsdfit.theory import GalaxyPowerParameters
--> 130 return GalaxyPowerParameters.from_defaults(model=self)
131
132 #---------------------------------------------------------------------------
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsdfit/theory/base.py in from_defaults(cls, model, extra_params)
155 """
156 # initialize an empty class
--> 157 params = cls()
158
159 # add extra parameters
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsdfit/parameters/parameterset.py in __init__(self, *args, **kwargs)
35 def __init__(self, *args, **kwargs):
36
---> 37 kwargs['asteval'] = lmfit.asteval.Interpreter(symtable=SmartSymTable(self))
38 super(ParameterSet, self).__init__(*args, **kwargs)
39 self._prepared = False
AttributeError: module 'lmfit' has no attribute 'asteval'
# fix the satellite fraction
In [4]: params['fs'].vary = False
---------------------------------------------------------------------------
NameError Traceback (most recent call last)
<ipython-input-4-7044e8003dc4> in <module>()
----> 1 params['fs'].vary = False
NameError: name 'params' is not defined
# write new configuration to a file
In [5]: params.to_file('params.dat', mode='a')
---------------------------------------------------------------------------
NameError Traceback (most recent call last)
<ipython-input-5-a71e8f6ee5f2> in <module>()
----> 1 params.to_file('params.dat', mode='a')
NameError: name 'params' is not defined
The Sample Number Density¶
The model requires the number density of the sample to properly account for 1-halo terms. This number density is assumed to be constant and in the case of a number density that varies with redshift, the value can be thought of an average number density.
Note
Typically, the number density value used is the inverse of the sample shot noise.
The value of the nbar parameter should be updated by the user, as
# get the default params
In [6]: params = model.default_params()
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-6-20168f551e3a> in <module>()
----> 1 params = model.default_params()
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsd/power/gal/power_gal.py in default_params(self)
128 """
129 from pyRSD.rsdfit.theory import GalaxyPowerParameters
--> 130 return GalaxyPowerParameters.from_defaults(model=self)
131
132 #---------------------------------------------------------------------------
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsdfit/theory/base.py in from_defaults(cls, model, extra_params)
155 """
156 # initialize an empty class
--> 157 params = cls()
158
159 # add extra parameters
~/checkouts/readthedocs.org/user_builds/pyrsd/conda/latest/lib/python3.6/site-packages/pyRSD/rsdfit/parameters/parameterset.py in __init__(self, *args, **kwargs)
35 def __init__(self, *args, **kwargs):
36
---> 37 kwargs['asteval'] = lmfit.asteval.Interpreter(symtable=SmartSymTable(self))
38 super(ParameterSet, self).__init__(*args, **kwargs)
39 self._prepared = False
AttributeError: module 'lmfit' has no attribute 'asteval'
# this is the default nbar value
In [7]: print(params['nbar'])
---------------------------------------------------------------------------
NameError Traceback (most recent call last)
<ipython-input-7-f96c9626798f> in <module>()
----> 1 print(params['nbar'])
NameError: name 'params' is not defined
# change to the right value
In [8]: params['nbar'].value = 4e-5 # in units of (Mpc/h)^{-3}
---------------------------------------------------------------------------
NameError Traceback (most recent call last)
<ipython-input-8-4ebd75800de5> in <module>()
----> 1 params['nbar'].value = 4e-5 # in units of (Mpc/h)^{-3}
NameError: name 'params' is not defined
Configuring the GalaxySpectrum Model¶
The GalaxySpectrum model has several configuration parameters,
which can be passed to __init__() function.
Most of these parameters have sensible default values that should not be
changed by the user. However, there are a few that should be changed based
on the data set being fit. The most important are the sample redshift and
the cosmological parameters.
The redshift of the sample is taken to be constant. For samples that have
a number density varying with redshift, the redshift should be set to the
effective redshift of the sample. In the parameter file passed to rsdfit
the redshift can be set as:
model.z = 0.55
The cosmological parameters can also be specified by the user in the
parameter file – they set the shape of the linear power spectrum which is the
input to the GalaxySpectrum model. However, if an
already initialized model is being passed to the rsdfit command via the
-m flag, then the user doesn’t necessarily need to specify the cosmology
again.
The cosmological parameters can be specified in a number of ways:
the name of a file
The cosmological parameters can be read from a parameter file. See the
pyRSD/data/paramsdirectory for example parameter files.the name of a builtin cosmology
The name of a builtin cosmology, such as
Planck15orWMAP9a dictionary of parameters
A dictionary of key/value pairs, such as
Ob0,Om0, etc, which will be passed to thepyRSD.rsd.cosmology.Cosmology.__init__function.
The cosmology parameters can be specified in the parameter file by
setting the model.params parameter. For example, to use the WMAP9
parameter set, simply specify the following in the parameter file:
model.params = "WMAP9"
Additionally, there are other model configuraton parameters that can be
set in the parameter file, as long as they are prefixed with model.. The
config attribute of the GalaxySpectrum object
gives these values
In [9]: from pyRSD.rsd import GalaxySpectrum
In [10]: model = GalaxySpectrum()
In [11]: for k in sorted(model.config):
....: print("%s = %s" %(k, str(model.config[k])))
....:
Nk = 200
Pdv_model_type = jennings
correct_mu2 = False
correct_mu4 = False
fog_model = modified_lorentzian
include_2loop = False
interpolate = True
k0_low = 0.005
kmax = 0.5
kmin = 0.001
linear_power_file = None
max_mu = 4
params = {'H0': 67.74, 'Om0': 0.3075, 'sigma8': 0.8159, 'n_s': 0.9667, 'flat': True, 'w0': -1.0, 'name': 'Planck15', 'Ob0': 0.0486, 'Tcmb0': 2.7255, 'Neff': 3.046, 'm_nu': array([0. , 0. , 0.06])}
transfer_fit = CLASS
use_P00_model = True
use_P01_model = True
use_P11_model = True
use_Pdv_model = True
use_Phm_model = True
use_Pvv_model = True
use_mean_bias = False
use_so_correction = False
use_tidal_bias = False
use_vlah_biasing = True
vel_disp_from_sims = False
z = 0.0