# Licensed under a 3-clause BSD style license - see LICENSE.rst
from pathlib import Path
import numpy as np
from astropy import units as u
from astropy.io import fits
from astropy.table import Table
from gammapy.data import GTI
from gammapy.irf import EDispKernel, EffectiveAreaTable, IRFStacker
from gammapy.modeling import Dataset, Parameters
from gammapy.modeling.models import Models
from gammapy.stats import cash, significance, significance_on_off, wstat
from gammapy.utils.fits import energy_axis_to_ebounds
from gammapy.utils.random import get_random_state
from gammapy.utils.scripts import make_name, make_path
from .core import CountsSpectrum, SpectrumEvaluator
__all__ = [
"SpectrumDatasetOnOff",
"SpectrumDataset",
"plot_spectrum_datasets_off_regions",
]
[docs]class SpectrumDataset(Dataset):
"""Spectrum dataset for likelihood fitting.
The spectrum dataset bundles reduced counts data, with a spectral model,
background model and instrument response function to compute the fit-statistic
given the current model and data.
Parameters
----------
models : `~gammapy.modeling.models.Models`
Fit model
counts : `~gammapy.spectrum.CountsSpectrum`
Counts spectrum
livetime : `~astropy.units.Quantity`
Livetime
aeff : `~gammapy.irf.EffectiveAreaTable`
Effective area
edisp : `~gammapy.irf.EDispKernel`
Energy dispersion
background : `~gammapy.spectrum.CountsSpectrum`
Background to use for the fit.
mask_safe : `~numpy.ndarray`
Mask defining the safe data range.
mask_fit : `~numpy.ndarray`
Mask to apply to the likelihood for fitting.
name : str
Dataset name.
gti : `~gammapy.data.GTI`
GTI of the observation or union of GTI if it is a stacked observation
See Also
--------
SpectrumDatasetOnOff, FluxPointsDataset, gammapy.cube.MapDataset
"""
stat_type = "cash"
tag = "SpectrumDataset"
def __init__(
self,
models=None,
counts=None,
livetime=None,
aeff=None,
edisp=None,
background=None,
mask_safe=None,
mask_fit=None,
name=None,
gti=None,
):
if mask_fit is not None and mask_fit.dtype != np.dtype("bool"):
raise ValueError("mask data must have dtype bool")
self.counts = counts
if livetime is not None:
livetime = u.Quantity(livetime)
self.livetime = livetime
self.mask_fit = mask_fit
self.aeff = aeff
self.edisp = edisp
self.background = background
self.models = models
self.mask_safe = mask_safe
self.gti = gti
self._name = make_name(name)
@property
def name(self):
return self._name
def __str__(self):
str_ = self.__class__.__name__ + "\n"
str_ += "-" * len(self.__class__.__name__) + "\n"
str_ += "\n"
str_ += "\t{:32}: {} \n\n".format("Name", self.name)
counts = np.nan
if self.counts is not None:
counts = np.sum(self.counts.data)
str_ += "\t{:32}: {:.0f} \n".format("Total counts", counts)
npred = np.nan
if self.models is not None:
npred = np.sum(self.npred().data)
str_ += "\t{:32}: {:.2f}\n".format("Total predicted counts", npred)
counts_off = np.nan
if getattr(self, "counts_off", None) is not None:
counts_off = np.sum(self.counts_off.data)
str_ += "\t{:32}: {:.2f}\n\n".format("Total off counts", counts_off)
background = np.nan
if getattr(self, "background", None) is not None:
background = np.sum(self.background.data)
str_ += "\t{:32}: {:.2f}\n\n".format("Total background counts", background)
aeff_min, aeff_max, aeff_unit = np.nan, np.nan, ""
if self.aeff is not None:
aeff_min = np.min(self.aeff.data.data.value[self.aeff.data.data.value > 0])
aeff_max = np.max(self.aeff.data.data.value)
aeff_unit = self.aeff.data.data.unit
str_ += "\t{:32}: {:.2e} {}\n".format("Effective area min", aeff_min, aeff_unit)
str_ += "\t{:32}: {:.2e} {}\n\n".format(
"Effective area max", aeff_max, aeff_unit
)
livetime = np.nan
if self.livetime is not None:
livetime = self.livetime
str_ += "\t{:32}: {:.2e}\n\n".format("Livetime", livetime)
# data section
n_bins = 0
if self.counts is not None:
n_bins = self.counts.data.size
str_ += "\t{:32}: {} \n".format("Number of total bins", n_bins)
n_fit_bins = 0
if self.mask is not None:
n_fit_bins = np.sum(self.mask)
str_ += "\t{:32}: {} \n\n".format("Number of fit bins", n_fit_bins)
# likelihood section
str_ += "\t{:32}: {}\n".format("Fit statistic type", self.stat_type)
stat = np.nan
if self.models is not None and self.counts is not None:
stat = self.stat_sum()
str_ += "\t{:32}: {:.2f}\n\n".format("Fit statistic value (-2 log(L))", stat)
n_pars, n_free_pars = 0, 0
if self.models is not None:
n_pars = len(self.models.parameters)
n_free_pars = len(self.parameters.free_parameters)
str_ += "\t{:32}: {}\n".format("Number of parameters", n_pars)
str_ += "\t{:32}: {}\n\n".format("Number of free parameters", n_free_pars)
if self.models is not None:
str_ += "\t" + "\n\t".join(str(self.models).split("\n")[2:])
return str_.expandtabs(tabsize=2)
@property
def models(self):
"""Models (`gammapy.modeling.models.Models`)."""
return self._models
@models.setter
def models(self, value):
if value is not None:
self._models = Models(value)
else:
self._models = None
evaluators = []
if self.models is not None:
for model in self.models:
evaluator = SpectrumEvaluator(
model=model,
livetime=self.livetime,
aeff=self.aeff,
edisp=self.edisp,
)
evaluators.append(evaluator)
self._evaluators = evaluators
@property
def mask_safe(self):
if self._mask_safe is None:
return np.ones(self.data_shape, bool)
else:
return self._mask_safe
@mask_safe.setter
def mask_safe(self, mask):
self._mask_safe = mask
@property
def parameters(self):
"""List of parameters (`~gammapy.modeling.Parameters`)"""
parameters = []
for component in self.models:
parameters.append(component.spectral_model.parameters)
return Parameters.from_stack(parameters)
@property
def _energy_axis(self):
if self.counts is not None:
e_axis = self.counts.energy
elif self.edisp is not None:
e_axis = self.edisp.data.axis("e_reco")
elif self.aeff is not None:
# assume e_reco = e_true
e_axis = self.aeff.data.axis("energy")
return e_axis
@property
def data_shape(self):
"""Shape of the counts data"""
return (self._energy_axis.nbin,)
[docs] def npred_sig(self):
"""Predicted counts from source model (`CountsSpectrum`)."""
data = np.zeros(self.data_shape)
npred = self._as_counts_spectrum(data)
for evaluator in self._evaluators:
npred += evaluator.compute_npred()
return npred
[docs] def npred(self):
"""Return npred map (model + background)"""
npred = self.npred_sig()
if self.background:
npred += self.background
return npred
[docs] def stat_array(self):
"""Likelihood per bin given the current model parameters"""
return cash(n_on=self.counts.data, mu_on=self.npred().data)
def _as_counts_spectrum(self, data):
energy = self._energy_axis.edges
return CountsSpectrum(data=data, energy_lo=energy[:-1], energy_hi=energy[1:])
@property
def excess(self):
"""Excess (counts - alpha * counts_off)"""
excess = self.counts.data - self.background.data
return self._as_counts_spectrum(excess)
[docs] def fake(self, random_state="random-seed", name=None):
"""Simulate fake counts for the current model and reduced irfs.
This method overwrites the counts defined on the dataset object.
Parameters
----------
random_state : {int, 'random-seed', 'global-rng', `~numpy.random.RandomState`}
Defines random number generator initialisation.
Passed to `~gammapy.utils.random.get_random_state`.
"""
self._name = make_name(name)
random_state = get_random_state(random_state)
npred = self.npred()
npred.data = random_state.poisson(npred.data)
self.counts = npred
@property
def energy_range(self):
"""Energy range defined by the safe mask"""
energy = self._energy_axis.edges
e_min, e_max = energy[:-1], energy[1:]
if self.mask_safe is not None:
if self.mask_safe.any():
e_min = e_min[self.mask_safe]
e_max = e_max[self.mask_safe]
else:
return None, None
return u.Quantity([e_min.min(), e_max.max()])
[docs] def plot_fit(self):
"""Plot counts and residuals in two panels.
Calls ``plot_counts`` and ``plot_residuals``.
"""
from matplotlib.gridspec import GridSpec
import matplotlib.pyplot as plt
gs = GridSpec(7, 1)
ax_spectrum = plt.subplot(gs[:5, :])
self.plot_counts(ax=ax_spectrum)
ax_spectrum.set_xticks([])
ax_residuals = plt.subplot(gs[5:, :])
self.plot_residuals(ax=ax_residuals)
return ax_spectrum, ax_residuals
@property
def _e_unit(self):
return self.counts.energy.unit
[docs] def plot_counts(self, ax=None):
"""Plot predicted and detected counts.
Parameters
----------
ax : `~matplotlib.pyplot.Axes`
Axes object.
Returns
-------
ax : `~matplotlib.pyplot.Axes`
Axes object.
"""
import matplotlib.pyplot as plt
ax = plt.gca() if ax is None else ax
self.npred_sig().plot(ax=ax, label="mu_src", energy_unit=self._e_unit)
self.excess.plot(ax=ax, label="Excess", fmt=".", energy_unit=self._e_unit)
e_min, e_max = self.energy_range
kwargs = {"color": "black", "linestyle": "dashed"}
ax.axvline(e_min.to_value(self._e_unit), label="fit range", **kwargs)
ax.axvline(e_max.to_value(self._e_unit), **kwargs)
ax.legend(numpoints=1)
ax.set_title("")
return ax
[docs] def residuals(self, method="diff"):
"""Compute the spectral residuals.
Parameters
----------
method : {"diff", "diff/model", "diff/sqrt(model)"}
Method used to compute the residuals. Available options are:
- ``diff`` (default): data - model
- ``diff/model``: (data - model) / model
- ``diff/sqrt(model)``: (data - model) / sqrt(model)
Returns
-------
residuals : `CountsSpectrum`
Residual spectrum
"""
residuals = self._compute_residuals(self.counts, self.npred(), method)
return residuals
[docs] def plot_residuals(self, method="diff", ax=None, **kwargs):
"""Plot residuals.
Parameters
----------
ax : `~matplotlib.pyplot.Axes`
Axes object.
method : {"diff", "diff/model", "diff/sqrt(model)"}
Normalization used to compute the residuals, see `SpectrumDataset.residuals()`
**kwargs : dict
Keywords passed to `CountsSpectrum.plot()`
Returns
-------
ax : `~matplotlib.pyplot.Axes`
Axes object.
"""
import matplotlib.pyplot as plt
ax = plt.gca() if ax is None else ax
residuals = self.residuals(method=method)
label = self._residuals_labels[method]
residuals.plot(
ax=ax, ecolor="black", fmt="none", energy_unit=self._e_unit, **kwargs
)
ax.axhline(0, color="black", lw=0.5)
ymax = 1.2 * np.nanmax(residuals.data)
ax.set_ylim(-ymax, ymax)
ax.set_xlabel(f"Energy [{self._e_unit}]")
ax.set_ylabel(f"Residuals ({label})")
return ax
[docs] @classmethod
def create(
cls, e_reco, e_true=None, region=None, reference_time="2000-01-01", name=None
):
"""Creates empty spectrum dataset.
Empty containers are created with the correct geometry.
counts, background and aeff are zero and edisp is diagonal.
The safe_mask is set to False in every bin.
Parameters
----------
e_reco : `~astropy.units.Quantity`
edges of counts vector
e_true : `~astropy.units.Quantity`
edges of effective area table. If not set use reco energy values. Default : None
region : `~regions.SkyRegion`
Region to define the dataset for.
reference_time : `~astropy.time.Time`
reference time of the dataset, Default is "2000-01-01"
"""
if e_true is None:
e_true = e_reco
counts = CountsSpectrum(e_reco[:-1], e_reco[1:], region=region)
background = CountsSpectrum(e_reco[:-1], e_reco[1:], region=region)
aeff = EffectiveAreaTable(
e_true[:-1], e_true[1:], np.zeros(e_true[:-1].shape) * u.m ** 2
)
edisp = EDispKernel.from_diagonal_response(e_true, e_reco)
mask_safe = np.zeros_like(counts.data, "bool")
gti = GTI.create(u.Quantity([], "s"), u.Quantity([], "s"), reference_time)
livetime = gti.time_sum
return SpectrumDataset(
counts=counts,
aeff=aeff,
edisp=edisp,
mask_safe=mask_safe,
background=background,
livetime=livetime,
gti=gti,
name=name,
)
[docs] def stack(self, other):
r"""Stack this dataset with another one.
Safe mask is applied to compute the stacked counts vector.
Counts outside each dataset safe mask are lost.
Stacking is performed in-place.
The stacking of 2 datasets is implemented as follows.
Here, :math:`k` denotes a bin in reconstructed energy and :math:`j = {1,2}` is the dataset number
The ``mask_safe`` of each dataset is defined as:
.. math::
\epsilon_{jk} =\left\{\begin{array}{cl} 1, &
\mbox{if bin k is inside the energy thresholds}\\ 0, &
\mbox{otherwise} \end{array}\right.
Then the total ``counts`` and model background ``bkg`` are computed according to:
.. math::
\overline{\mathrm{n_{on}}}_k = \mathrm{n_{on}}_{1k} \cdot \epsilon_{1k} +
\mathrm{n_{on}}_{2k} \cdot \epsilon_{2k}
\overline{bkg}_k = bkg_{1k} \cdot \epsilon_{1k} +
bkg_{2k} \cdot \epsilon_{2k}
The stacked ``safe_mask`` is then:
.. math::
\overline{\epsilon_k} = \epsilon_{1k} OR \epsilon_{2k}
Please refer to the `~gammapy.irf.IRFStacker` for the description
of how the IRFs are stacked.
Parameters
----------
other : `~gammapy.spectrum.SpectrumDataset`
the dataset to stack to the current one
"""
if not isinstance(other, SpectrumDataset):
raise TypeError("Incompatible types for SpectrumDataset stacking")
if self.counts is not None:
self.counts.data[~self.mask_safe] = 0
self.counts.data[other.mask_safe] += other.counts.data[other.mask_safe]
if self.background is not None:
self.background.data[~self.mask_safe] = 0
self.background.data[other.mask_safe] += other.background.data[
other.mask_safe
]
if self.aeff is not None:
if self.livetime is None or other.livetime is None:
raise ValueError("IRF stacking requires livetime for both datasets.")
irf_stacker = IRFStacker(
list_aeff=[self.aeff, other.aeff],
list_livetime=[self.livetime, other.livetime],
list_edisp=[self.edisp, other.edisp],
list_low_threshold=[self.energy_range[0], other.energy_range[0]],
list_high_threshold=[self.energy_range[1], other.energy_range[1]],
)
irf_stacker.stack_aeff()
if self.edisp is not None:
irf_stacker.stack_edisp()
self.edisp = irf_stacker.stacked_edisp
self.aeff = irf_stacker.stacked_aeff
self.mask_safe = np.logical_or(self.mask_safe, other.mask_safe)
if self.gti is not None:
self.gti = self.gti.stack(other.gti).union()
# TODO: for the moment, since dead time is not accounted for, livetime cannot be the sum of GTIs
if self.livetime is not None:
self.livetime += other.livetime
[docs] def peek(self, figsize=(16, 4)):
"""Quick-look summary plots."""
import matplotlib.pyplot as plt
e_min, e_max = self.energy_range
_, (ax1, ax2, ax3) = plt.subplots(nrows=1, ncols=3, figsize=figsize)
ax1.set_title("Counts")
energy_unit = "TeV"
if isinstance(self, SpectrumDatasetOnOff) and self.counts_off is not None:
self.background.plot_hist(
ax=ax1, label="alpha * N_off", color="darkblue", energy_unit=energy_unit
)
elif self.background is not None:
self.background.plot_hist(
ax=ax1, label="background", color="darkblue", energy_unit=energy_unit
)
self.counts.plot_hist(
ax=ax1,
label="n_on",
color="darkred",
energy_unit=energy_unit,
show_energy=(e_min, e_max),
)
ax1.set_xlim(
0.7 * e_min.to_value(energy_unit), 1.3 * e_max.to_value(energy_unit)
)
ax1.legend(numpoints=1)
ax2.set_title("Effective Area")
e_unit = self.aeff.energy.unit
self.aeff.plot(ax=ax2, show_energy=(e_min, e_max))
ax2.set_xlim(0.7 * e_min.to_value(e_unit), 1.3 * e_max.to_value(e_unit))
ax3.set_title("Energy Dispersion")
if self.edisp is not None:
self.edisp.plot_matrix(ax=ax3)
# TODO: optimize layout
plt.subplots_adjust(wspace=0.3)
[docs] def info_dict(self, in_safe_energy_range=True):
"""Info dict with summary statistics, summed over energy
Parameters
----------
in_safe_energy_range : bool
Whether to sum only in the safe energy range
Returns
-------
info_dict : dict
Dictionary with summary info.
"""
info = dict()
mask = self.mask_safe if in_safe_energy_range else slice(None)
info["name"] = self.name
info["livetime"] = self.livetime.copy()
info["n_on"] = self.counts.data[mask].sum()
info["background"] = self.background.data[mask].sum()
info["excess"] = self.excess.data[mask].sum()
info["significance"] = significance(
self.counts.data[mask].sum(), self.background.data[mask].sum(),
)
info["background_rate"] = info["background"] / info["livetime"]
info["gamma_rate"] = info["excess"] / info["livetime"]
return info
[docs]class SpectrumDatasetOnOff(SpectrumDataset):
"""Spectrum dataset for on-off likelihood fitting.
The on-off spectrum dataset bundles reduced counts data, off counts data,
with a spectral model, relative background efficiency and instrument
response functions to compute the fit-statistic given the current model
and data.
Parameters
----------
models : `~gammapy.modeling.models.Models`
Fit model
counts : `~gammapy.spectrum.CountsSpectrum`
ON Counts spectrum
counts_off : `~gammapy.spectrum.CountsSpectrum`
OFF Counts spectrum
livetime : `~astropy.units.Quantity`
Livetime
aeff : `~gammapy.irf.EffectiveAreaTable`
Effective area
edisp : `~gammapy.irf.EDispKernel`
Energy dispersion
mask_safe : `~numpy.array`
Mask defining the safe data range.
mask_fit : `~numpy.array`
Mask to apply to the likelihood for fitting.
acceptance : `~numpy.array` or float
Relative background efficiency in the on region.
acceptance_off : `~numpy.array` or float
Relative background efficiency in the off region.
name : str
Name of the dataset.
gti : `~gammapy.data.GTI`
GTI of the observation or union of GTI if it is a stacked observation
See Also
--------
SpectrumDataset, FluxPointsDataset, MapDataset
"""
stat_type = "wstat"
tag = "SpectrumDatasetOnOff"
def __init__(
self,
models=None,
counts=None,
counts_off=None,
livetime=None,
aeff=None,
edisp=None,
mask_safe=None,
mask_fit=None,
acceptance=None,
acceptance_off=None,
name=None,
gti=None,
):
self.counts = counts
self.counts_off = counts_off
if livetime is not None:
livetime = u.Quantity(livetime)
self.livetime = livetime
self.mask_fit = mask_fit
self.aeff = aeff
self.edisp = edisp
self.models = models
self.mask_safe = mask_safe
if np.isscalar(acceptance):
acceptance = np.ones(self.data_shape) * acceptance
if np.isscalar(acceptance_off):
acceptance_off = np.ones(self.data_shape) * acceptance_off
self.acceptance = acceptance
self.acceptance_off = acceptance_off
self._name = make_name(name)
self.gti = gti
def __str__(self):
str_ = super().__str__()
str_list = str_.split("\n")
acceptance = np.nan
if self.acceptance is not None:
acceptance = np.mean(self.acceptance)
str_acc = "\t{:32}: {}\n".format("Acceptance mean:", acceptance)
str_list.insert(16, str_acc)
str_ = "\n".join(str_list)
return str_.expandtabs(tabsize=2)
@property
def background(self):
""""""
background = self.alpha * self.counts_off.data
return self._as_counts_spectrum(background)
@property
def alpha(self):
"""Exposure ratio between signal and background regions"""
return self.acceptance / self.acceptance_off
[docs] def stat_array(self):
"""Likelihood per bin given the current model parameters"""
mu_sig = self.npred_sig().data
on_stat_ = wstat(
n_on=self.counts.data,
n_off=self.counts_off.data,
alpha=self.alpha,
mu_sig=mu_sig,
)
return np.nan_to_num(on_stat_)
[docs] def fake(self, background_model, random_state="random-seed", name=None):
"""Simulate fake counts for the current model and reduced irfs.
This method overwrites the counts and off counts defined on the dataset object.
Parameters
----------
background_model : `~gammapy.spectrum.CountsSpectrum`
BackgroundModel. In the future will be part of the SpectrumDataset Class.
For the moment, a CountSpectrum.
random_state : {int, 'random-seed', 'global-rng', `~numpy.random.RandomState`}
Defines random number generator initialisation.
Passed to `~gammapy.utils.random.get_random_state`.
"""
self._name = make_name(name)
random_state = get_random_state(random_state)
npred_sig = self.npred_sig()
npred_sig.data = random_state.poisson(npred_sig.data)
npred_bkg = background_model.copy()
npred_bkg.data = random_state.poisson(npred_bkg.data)
self.counts = npred_sig + npred_bkg
npred_off = background_model / self.alpha
npred_off.data = random_state.poisson(npred_off.data)
self.counts_off = npred_off
[docs] @classmethod
def create(
cls, e_reco, e_true=None, region=None, reference_time="2000-01-01", name=None
):
"""Create empty SpectrumDatasetOnOff.
Empty containers are created with the correct geometry.
counts, counts_off and aeff are zero and edisp is diagonal.
The safe_mask is set to False in every bin.
Parameters
----------
e_reco : `~astropy.units.Quantity`
edges of counts vector
e_true : `~astropy.units.Quantity`
edges of effective area table. If not set use reco energy values. Default : None
region : `~regions.SkyRegion`
Region to define the dataset for.
reference_time : `~astropy.time.Time`
reference time of the dataset, Default is "2000-01-01"
"""
if e_true is None:
e_true = e_reco
counts = CountsSpectrum(e_reco[:-1], e_reco[1:], region=region)
counts_off = CountsSpectrum(e_reco[:-1], e_reco[1:], region=region)
aeff = EffectiveAreaTable(
e_true[:-1], e_true[1:], np.zeros(e_true[:-1].shape) * u.m ** 2
)
edisp = EDispKernel.from_diagonal_response(e_true, e_reco)
mask_safe = np.zeros_like(counts.data, "bool")
gti = GTI.create(u.Quantity([], "s"), u.Quantity([], "s"), reference_time)
livetime = gti.time_sum
acceptance = np.ones_like(counts.data, int)
acceptance_off = np.ones_like(counts.data, int)
return SpectrumDatasetOnOff(
counts=counts,
counts_off=counts_off,
aeff=aeff,
edisp=edisp,
mask_safe=mask_safe,
acceptance=acceptance,
acceptance_off=acceptance_off,
livetime=livetime,
gti=gti,
name=name,
)
[docs] @classmethod
def read(cls, filename):
"""Read from file
For now, filename is assumed to the name of a PHA file where BKG file, ARF, and RMF names
must be set in the PHA header and be present in the same folder
Parameters
----------
filename : str
OGIP PHA file to read
"""
raise NotImplementedError(
"To read from an OGIP fits file use SpectrumDatasetOnOff.from_ogip_files."
)
def _is_stackable(self):
"""Check if the Dataset contains enough information to be stacked"""
if (
self.acceptance_off is None
or self.acceptance is None
or self.counts_off is None
):
return False
else:
return True
[docs] def stack(self, other):
r"""Stack this dataset with another one.
Safe mask is applied to compute the stacked counts vector.
Counts outside each dataset safe mask are lost.
Stacking is performed in-place.
The stacking of 2 datasets is implemented as follows.
Here, :math:`k` denotes a bin in reconstructed energy and :math:`j = {1,2}` is the dataset number
The ``mask_safe`` of each dataset is defined as:
.. math::
\epsilon_{jk} =\left\{\begin{array}{cl} 1, &
\mbox{if k is inside the energy thresholds}\\ 0, &
\mbox{otherwise} \end{array}\right.
Then the total ``counts`` and ``counts_off`` are computed according to:
.. math::
\overline{\mathrm{n_{on}}}_k = \mathrm{n_{on}}_{1k} \cdot \epsilon_{1k} +
\mathrm{n_{on}}_{2k} \cdot \epsilon_{2k}
\overline{\mathrm{n_{off}}}_k = \mathrm{n_{off}}_{1k} \cdot \epsilon_{1k} +
\mathrm{n_{off}}_{2k} \cdot \epsilon_{2k}
The stacked ``safe_mask`` is then:
.. math::
\overline{\epsilon_k} = \epsilon_{1k} OR \epsilon_{2k}
In each energy bin :math:`k`, the count excess is computed taking into account the ON ``acceptance``,
:math:`a_{on}_k` and the OFF one: ``acceptance_off``, :math:`a_{off}_k`. They define
the :math:`\alpha_k=a_{on}_k/a_{off}_k` factors such that :math:`n_{ex}_k = n_{on}_k - \alpha_k n_{off}_k`.
We define the stacked value of :math:`\overline{{a}_{on}}_k = 1` so that:
.. math::
\overline{{a}_{off}}_k = \frac{\overline{\mathrm {n_{off}}}}{\alpha_{1k} \cdot
\mathrm{n_{off}}_{1k} \cdot \epsilon_{1k} + \alpha_{2k} \cdot \mathrm{n_{off}}_{2k} \cdot \epsilon_{2k}}
Please refer to the `~gammapy.irf.IRFStacker` for the description
of how the IRFs are stacked.
Parameters
----------
other : `~gammapy.spectrum.SpectrumDatasetOnOff`
the dataset to stack to the current one
Examples
--------
>>> from gammapy.spectrum import SpectrumDatasetOnOff
>>> obs_ids = [23523, 23526, 23559, 23592]
>>> datasets = []
>>> for obs in obs_ids:
>>> filename = "$GAMMAPY_DATA/joint-crab/spectra/hess/pha_obs{}.fits"
>>> ds = SpectrumDatasetOnOff.from_ogip_files(filename.format(obs))
>>> datasets.append(ds)
>>> stacked = datasets[0]
>>> for ds in datasets[1:]:
>>> stacked.stack(ds)
>>> print(stacked.livetime)
6313.8116406202325 s
"""
if not isinstance(other, SpectrumDatasetOnOff):
raise TypeError("Incompatible types for SpectrumDatasetOnOff stacking")
# We assume here that counts_off, acceptance and acceptance_off are well defined.
if not self._is_stackable() or not other._is_stackable():
raise ValueError("Cannot stack incomplete SpectrumDatsetOnOff.")
total_off = np.zeros_like(self.counts_off.data, dtype=float)
total_alpha = np.zeros_like(self.counts_off.data, dtype=float)
total_off[self.mask_safe] += self.counts_off.data[self.mask_safe]
total_off[other.mask_safe] += other.counts_off.data[other.mask_safe]
total_alpha[self.mask_safe] += (self.alpha * self.counts_off)[self.mask_safe]
total_alpha[other.mask_safe] += (other.alpha * other.counts_off)[
other.mask_safe
]
with np.errstate(divide="ignore", invalid="ignore"):
acceptance_off = total_off / total_alpha
average_alpha = total_alpha.sum() / total_off.sum()
# For the bins where the stacked OFF counts equal 0, the alpha value is performed by weighting on the total
# OFF counts of each run
is_zero = total_off == 0
acceptance_off[is_zero] = 1 / average_alpha
self.acceptance = np.ones_like(self.counts_off.data, dtype=float)
self.acceptance_off = acceptance_off
if self.counts_off is not None:
self.counts_off.data[~self.mask_safe] = 0
self.counts_off.data[other.mask_safe] += other.counts_off.data[
other.mask_safe
]
super().stack(other)
[docs] def to_ogip_files(self, outdir=None, use_sherpa=False, overwrite=False):
"""Write OGIP files.
If you want to use the written files with Sherpa you have to set the
``use_sherpa`` flag. Then all files will be written in units 'keV' and
'cm2'.
The naming scheme is fixed, with {name} the dataset name:
* PHA file is named pha_obs{name}.fits
* BKG file is named bkg_obs{name}.fits
* ARF file is named arf_obs{name}.fits
* RMF file is named rmf_obs{name}.fits
Parameters
----------
outdir : `pathlib.Path`
output directory, default: pwd
use_sherpa : bool, optional
Write Sherpa compliant files, default: False
overwrite : bool
Overwrite existing files?
"""
# TODO: refactor and reduce amount of code duplication
outdir = Path.cwd() if outdir is None else make_path(outdir)
outdir.mkdir(exist_ok=True, parents=True)
phafile = f"pha_obs{self.name}.fits"
bkgfile = phafile.replace("pha", "bkg")
arffile = phafile.replace("pha", "arf")
rmffile = phafile.replace("pha", "rmf")
counts_table = self.counts.to_table()
counts_table["QUALITY"] = np.logical_not(self.mask_safe)
counts_table["BACKSCAL"] = self.acceptance
counts_table["AREASCAL"] = np.ones(self.acceptance.size)
meta = self._ogip_meta()
meta["respfile"] = rmffile
meta["backfile"] = bkgfile
meta["ancrfile"] = arffile
meta["hduclas2"] = "TOTAL"
counts_table.meta = meta
name = counts_table.meta["name"]
hdu = fits.BinTableHDU(counts_table, name=name)
hdulist = fits.HDUList([fits.PrimaryHDU(), hdu, self._ebounds_hdu(use_sherpa)])
if self.gti is not None:
hdu = fits.BinTableHDU(self.gti.table, name="GTI")
hdulist.append(hdu)
hdulist.writeto(outdir / phafile, overwrite=overwrite)
self.aeff.write(outdir / arffile, overwrite=overwrite, use_sherpa=use_sherpa)
if self.counts_off is not None:
counts_off_table = self.counts_off.to_table()
counts_off_table["QUALITY"] = np.logical_not(self.mask_safe)
counts_off_table["BACKSCAL"] = self.acceptance_off
counts_off_table["AREASCAL"] = np.ones(self.acceptance.size)
meta = self._ogip_meta()
meta["hduclas2"] = "BKG"
counts_off_table.meta = meta
name = counts_off_table.meta["name"]
hdu = fits.BinTableHDU(counts_off_table, name=name)
hdulist = fits.HDUList(
[fits.PrimaryHDU(), hdu, self._ebounds_hdu(use_sherpa)]
)
hdulist.writeto(outdir / bkgfile, overwrite=overwrite)
if self.edisp is not None:
self.edisp.write(
outdir / rmffile, overwrite=overwrite, use_sherpa=use_sherpa
)
def _ebounds_hdu(self, use_sherpa):
energy = self.counts.energy.edges
if use_sherpa:
energy = energy.to("keV")
return energy_axis_to_ebounds(energy)
def _ogip_meta(self):
"""Meta info for the OGIP data format"""
return {
"name": "SPECTRUM",
"hduclass": "OGIP",
"hduclas1": "SPECTRUM",
"corrscal": "",
"chantype": "PHA",
"detchans": self.counts.energy.nbin,
"filter": "None",
"corrfile": "",
"poisserr": True,
"hduclas3": "COUNT",
"hduclas4": "TYPE:1",
"lo_thres": self.energy_range[0].to_value("TeV"),
"hi_thres": self.energy_range[1].to_value("TeV"),
"exposure": self.livetime.to_value("s"),
"obs_id": self.name,
}
[docs] @classmethod
def from_ogip_files(cls, filename):
"""Read `~gammapy.spectrum.SpectrumDatasetOnOff` from OGIP files.
BKG file, ARF, and RMF must be set in the PHA header and be present in
the same folder.
The naming scheme is fixed to the following scheme:
* PHA file is named ``pha_obs{name}.fits``
* BKG file is named ``bkg_obs{name}.fits``
* ARF file is named ``arf_obs{name}.fits``
* RMF file is named ``rmf_obs{name}.fits``
with ``{name}`` the dataset name.
Parameters
----------
filename : str
OGIP PHA file to read
"""
filename = make_path(filename)
dirname = filename.parent
with fits.open(filename, memmap=False) as hdulist:
data = _read_ogip_hdulist(hdulist)
counts = CountsSpectrum(
energy_hi=data["energy_hi"], energy_lo=data["energy_lo"], data=data["data"]
)
phafile = filename.name
try:
rmffile = phafile.replace("pha", "rmf")
energy_dispersion = EDispKernel.read(dirname / rmffile)
except OSError:
# TODO : Add logger and echo warning
energy_dispersion = None
try:
bkgfile = phafile.replace("pha", "bkg")
with fits.open(dirname / bkgfile, memmap=False) as hdulist:
data_bkg = _read_ogip_hdulist(hdulist)
counts_off = CountsSpectrum(
energy_hi=data_bkg["energy_hi"],
energy_lo=data_bkg["energy_lo"],
data=data_bkg["data"],
)
acceptance_off = data_bkg["backscal"]
except OSError:
# TODO : Add logger and echo warning
counts_off, acceptance_off = None, None
arffile = phafile.replace("pha", "arf")
aeff = EffectiveAreaTable.read(dirname / arffile)
mask_safe = np.logical_not(data["quality"])
return cls(
counts=counts,
aeff=aeff,
counts_off=counts_off,
edisp=energy_dispersion,
livetime=data["livetime"],
mask_safe=mask_safe,
acceptance=data["backscal"],
acceptance_off=acceptance_off,
name=str(data["obs_id"]),
gti=data["gti"],
)
[docs] def info_dict(self, in_safe_energy_range=True):
"""Info dict with summary statistics, summed over energy
Parameters
----------
in_safe_energy_range : bool
Whether to sum only in the safe energy range
Returns
-------
info_dict : dict
Dictionary with summary info.
"""
info = super().info_dict(in_safe_energy_range)
mask = self.mask_safe if in_safe_energy_range else slice(None)
# TODO: handle energy dependent a_on / a_off
info["a_on"] = self.acceptance[0].copy()
if self.counts_off is not None:
info["n_off"] = self.counts_off.data[mask].sum()
info["a_off"] = self.acceptance_off[0].copy()
else:
info["n_off"] = 0
info["a_off"] = 1
info["alpha"] = self.alpha[0].copy()
info["significance"] = significance_on_off(
self.counts.data[mask].sum(),
self.counts_off.data[mask].sum(),
self.alpha[0],
)
return info
[docs] def to_dict(self, filename, *args, **kwargs):
"""Convert to dict for YAML serialization."""
outdir = Path(filename).parent
filename = str(outdir / f"pha_obs{self.name}.fits")
if self.models is None:
models = []
else:
models = [_.name for _ in self.models]
return {
"name": self.name,
"type": self.tag,
"models": models,
"filename": filename,
}
[docs] def write(self, filename, overwrite):
"""Write spectrum dataset on off to file.
Currently only the OGIP format is supported
Parameters
----------
filename : str
Filename to write to.
overwrite : bool
Overwrite existing file.
"""
outdir = Path(filename).parent
self.to_ogip_files(outdir=outdir, overwrite=overwrite)
[docs] @classmethod
def from_dict(cls, data, components, models):
"""Create flux point dataset from dict.
Parameters
----------
data : dict
Dict containing data to create dataset from.
components : list of dict
Not used.
models : list of `SkyModel`
List of model components.
Returns
-------
dataset : `SpectrumDatasetOnOff`
Spectrum dataset on off.
"""
models = [model for model in models if model.name in data["models"]]
# TODO: assumes that the model is a skymodel
# so this will work only when this change will be effective
filename = data["filename"]
dataset = cls.from_ogip_files(filename=filename)
dataset.mask_fit = None
dataset.models = models
return dataset
[docs] @classmethod
def from_spectrum_dataset(
cls, dataset, acceptance, acceptance_off, counts_off=None
):
"""Create spectrum dataseton off from another dataset.
Parameters
----------
dataset : `SpectrumDataset`
Spectrum dataset defining counts, edisp, aeff, livetime etc.
acceptance : `~numpy.array` or float
Relative background efficiency in the on region.
acceptance_off : `~numpy.array` or float
Relative background efficiency in the off region.
counts_off : `~gammapy.spectrum.CountsSpectrum`
Off counts spectrum . If the dataset provides a background model,
and no off counts are defined. The off counts are deferred from
counts_off / alpha.
Returns
-------
dataset : `SpectrumDatasetOnOff`
Spectrum dataset on off.
"""
if counts_off is None and dataset.background is not None:
alpha = acceptance / acceptance_off
counts_off = dataset.background / alpha
return cls(
counts=dataset.counts,
aeff=dataset.aeff,
counts_off=counts_off,
edisp=dataset.edisp,
livetime=dataset.livetime,
mask_safe=dataset.mask_safe,
mask_fit=dataset.mask_fit,
acceptance=acceptance,
acceptance_off=acceptance_off,
gti=dataset.gti,
name=dataset.name,
)
def _read_ogip_hdulist(hdulist, hdu1="SPECTRUM", hdu2="EBOUNDS", hdu3="GTI"):
"""Create from `~astropy.io.fits.HDUList`."""
counts_table = Table.read(hdulist[hdu1])
ebounds = Table.read(hdulist[hdu2])
emin = ebounds["E_MIN"].quantity
emax = ebounds["E_MAX"].quantity
if hdu3 in hdulist:
gti = GTI(Table.read(hdulist[hdu3]))
else:
gti = None
# Check if column are present in the header
quality = None
areascal = None
backscal = None
if "QUALITY" in counts_table.colnames:
quality = counts_table["QUALITY"].data
if "AREASCAL" in counts_table.colnames:
areascal = counts_table["AREASCAL"].data
if "BACKSCAL" in counts_table.colnames:
backscal = counts_table["BACKSCAL"].data
return dict(
data=counts_table["COUNTS"],
backscal=backscal,
energy_lo=emin,
energy_hi=emax,
quality=quality,
areascal=areascal,
livetime=counts_table.meta["EXPOSURE"] * u.s,
obs_id=counts_table.meta["OBS_ID"],
is_bkg=False,
gti=gti,
)
[docs]def plot_spectrum_datasets_off_regions(datasets, ax=None):
"""Plot spectrum datasets of regions.
Parameters
----------
datasets : list of `SpectrumDatasetOnOff`
List of spectrum on-off datasets
"""
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
ax = plt.gca() or ax
color_cycle = plt.rcParams["axes.prop_cycle"]
colors = color_cycle.by_key()["color"]
handles = []
for color, dataset in zip(colors, datasets):
kwargs = {"edgecolor": color, "facecolor": "none"}
dataset.counts_off.plot_region(ax=ax, **kwargs)
# create proxy artist for the custom legend
handle = mpatches.Patch(label=dataset.name, **kwargs)
handles.append(handle)
plt.legend(handles=handles)