def get_1d_halo_matter(self, kind='small_r', average='box'):
"""
Return the 1D real-space halo-matter correlation functions,
with errors estimated from the scatter in the 10 realizations
"""
valid_kinds = ['small_r', 'fft']
if kind not in valid_kinds:
raise ValueError("valid choices for `kind` are: %s" %str(valid_kinds))
name = '_1d_halo_matter_%s' %kind
if average is not None:
name += '_'+average
try:
return getattr(self, name)
except AttributeError:
basename = 'corr_1d_hm{mass}_runPB_PB{box}_{a}_%s.dat' %kind
d = os.path.join(self.root, 'halo-matter/config/real')
loader = io.load_correlation
coords = [self.a, self.mass, self.box]; dims = ['a', 'mass', 'box']
corr = SpectraSet.from_files(loader, d, basename, coords, dims, args=('1d',))
# reindex and add the errors
corr = self.reindex(corr, 'r_cen', self.dr, weights='N')
# compute errors
errs = {}
for key in corr.ndindex(dims=['a', 'mass']):
xi = corr.sel(**key)
data = []
for x in xi:
x = x.get()
data.append(x['corr'])
subkey = (key['a'], key['mass'])
errs[subkey] = np.diag(np.cov(np.asarray(data).T))**0.5
if average is not None:
errs[subkey] /= (len(self.box))**0.5
# average?
if average is not None:
corr = corr.average(axis=average)
# add the errors
for key in corr.ndindex():
xi = corr.loc[key].get()
xi['error'] = errs[(key['a'], key['mass'])]
setattr(self, name, corr)
return corr
def get_subbox_poles(self, which, average=False, subtract_shot_noise=True):
"""
Return the cutsky galaxy multipoles in redshift space, measured
for the subboxes
"""
name = '_my_poles_subboxes_%s_Mpch' %(which)
if average: name += '_mean'
try:
return getattr(self, name)
except AttributeError:
# form the filename and load the data
d = self.root
basename = 'poles_my_cutskyN1_subbox{box:d}_%s_Mpch_unscaled_no_fkp_dk005.dat' %which
# read in the data
loader = io.load_power
mapcols = {'power_0.real':'mono', 'power_2.real':'quad', 'power_4.real':'hexadec'}
usecols = ['k', 'mono', 'quad', 'hexadec', 'modes']
kwargs = {'usecols':usecols, 'mapcols':mapcols}
if which == '300':
boxes = list(range(94))
else:
boxes = list(range(24))
poles = SpectraSet.from_files(loader, d, basename, [boxes], ['box'], args=('1d',), kwargs=kwargs, ignore_missing=True)
# remove null
valid_boxes = []
for key, p in poles.nditer():
if not p.isnull():
valid_boxes.append(key['box'])
poles = poles.sel(box=valid_boxes)
# reindex
poles = self.reindex(poles, 'k_cen', self.dk, weights='modes')
# unstack the poles
ells = [('mono',0), ('quad', 2), ('hexadec', 4)]
poles = tools.unstack_multipoles(poles, ells, 'power')
# compute errors
errs = {}
for ell in poles['ell'].values:
data = []
for box in poles['box'].values:
p = poles.sel(box=box, ell=ell).get()
data.append(p['power'])
errs[ell] = np.diag(np.cov(np.asarray(data).T))**0.5
if average:
errs[ell] /= (len(boxes))**0.5
if subtract_shot_noise:
for key in poles.ndindex():
if key['ell'] == 0:
p = poles.loc[key].get()
p['power'] = p['power'] - p.attrs['shot_noise']
# average?
if average:
poles = poles.average(axis='box')
# add the errors
for key in poles.ndindex():
p = poles.loc[key].get()
p['error'] = errs[key['ell']]
setattr(self, name, poles)
return poles
def get_my_poles(self, space='redshift', scaled=False, average=False, tag="",
subtract_shot_noise=True):
"""
Return the cutsky galaxy multipoles in redshift space, measured
for my 84 box realizations
"""
if space not in ['real', 'redshift']:
raise ValueError("`space` should be 'real' or 'redshift'")
scaled_tag = 'scaled' if scaled else 'unscaled'
name = '_my_poles_%s_%s' %(space, scaled_tag)
if average: name += '_mean'
if tag:
tag = '_'+tag
name += tag
try:
return getattr(self, name)
except AttributeError:
# form the filename and load the data
d = os.path.join(self.root, 'nbodykit/poles')
if space == 'redshift':
basename = 'poles_my_cutskyN{box:d}_%s_no_fkp_dk005%s.dat' %(scaled_tag, tag)
else:
basename = 'poles_my_cutskyN{box:d}_real_%s_no_fkp_dk005%s.dat' %(scaled_tag, tag)
# read in the data
loader = io.load_power
mapcols = {'power_0.real':'mono', 'power_2.real':'quad', 'power_4.real':'hexadec'}
usecols = ['k', 'mono', 'quad', 'hexadec', 'modes']
kwargs = {'usecols':usecols, 'mapcols':mapcols}
poles = SpectraSet.from_files(loader, d, basename, [self.boxes], ['box'], args=('1d',), kwargs=kwargs, ignore_missing=True)
# remove null
valid_boxes = []
for key, p in poles.nditer():
if not p.isnull():
valid_boxes.append(key['box'])
poles = poles.sel(box=valid_boxes)
# reindex
poles = self.reindex(poles, 'k_cen', self.dk, weights='modes')
# unstack the poles
ells = [('mono',0), ('quad', 2), ('hexadec', 4)]
poles = tools.unstack_multipoles(poles, ells, 'power')
# compute errors
errs = {}
for ell in poles['ell'].values:
data = []
for box in poles['box'].values:
p = poles.sel(box=box, ell=ell).get()
data.append(p['power'])
errs[ell] = np.diag(np.cov(np.asarray(data).T))**0.5
if average:
errs[ell] /= (len(self.boxes))**0.5
if subtract_shot_noise:
for key in poles.ndindex():
if key['ell'] == 0:
p = poles.loc[key].get()
p['power'] = p['power'] - p.attrs['shot_noise']
# average?
if average:
poles = poles.average(axis='box')
# add the errors
for key in poles.ndindex():
p = poles.loc[key].get()
p['error'] = errs[key['ell']]
setattr(self, name, poles)
return poles
def _errors_from_realizations(self, a, mass, moments, sel_mu=None):
"""
Internal function to compute the momentum moment errors as the
diagonal elements of the covariance matrix, which is constructed using
the 30 different realizations of each spectrum
(3 lines-of-sight + 10 realizations)
Parameters
----------
a : str
the string specifying the scale factor
mass : int
the integer specifying the mass bin
moments : list of tuples
list of tuples specifying the velocity moments to load
"""
from glob import glob
from collections import defaultdict
if isinstance(sel_mu, int): sel_mu = [sel_mu]
if sel_mu is None: sel_mu = self.mu
Nmu = len(self.mu)
# the file pattern
d = "/Volumes/Frodo/RSDResearch/RunPB/Results/halo/real/momentum/poles"
data = defaultdict(list)
for i, (ell, ell_prime) in enumerate(moments):
args = (ell, ell_prime, mass, a)
basename = 'poles_P%d%d_hh%d_runPB_*_%s_*los.dat' %args
pattern = os.path.join(d, basename)
# the matching files
files = glob(pattern)
kw = {'sum_only':['modes'], 'force_index_match':True}
for j, f in enumerate(files):
loaded = io.load_momentum(f, ell, ell_prime, **kw)
if self.dk is not None:
for ii, l in enumerate(loaded):
if isinstance(l, DataSet):
loaded[ii] = l.reindex('k_cen', self.dk, weights='modes')
for k, mu in enumerate(self.mu):
if mu not in sel_mu: continue
if isinstance(loaded[k], DataSet):
if i == 0:
data[mu].append(loaded[k]['power'])
else:
data[mu][j] += loaded[k]['power']
toret = [None]*Nmu
for i, mu in enumerate(self.mu):
if len(data[mu]):
cov = np.cov(np.nan_to_num(data[mu]), rowvar=False)
errs = cov.diagonal()**0.5
# divide by root N if mean
if 'mean' in self.tag:
errs /= len(files)**0.5
toret[i] = errs
return toret
