def make_observation_list():
"""obs with dummy IRF"""
nbin = 3
energy = np.logspace(-1, 1, nbin + 1) * u.TeV
livetime = 2 * u.h
data_on = np.arange(nbin)
dataoff_1 = np.ones(3)
dataoff_2 = np.ones(3) * 3
dataoff_1[1] = 0
dataoff_2[1] = 0
on_vector = CountsSpectrum(energy_lo=energy[:-1],
energy_hi=energy[1:],
data=data_on)
off_vector1 = CountsSpectrum(energy_lo=energy[:-1],
energy_hi=energy[1:],
data=dataoff_1)
off_vector2 = CountsSpectrum(energy_lo=energy[:-1],
energy_hi=energy[1:],
data=dataoff_2)
aeff = EffectiveAreaTable.from_constant(energy, "1 cm2")
edisp = EDispKernel.from_gauss(e_true=energy,
e_reco=energy,
sigma=0.2,
bias=0)
time_ref = Time("2010-01-01")
gti1 = make_gti({
"START": [5, 6, 1, 2],
"STOP": [8, 7, 3, 4]
},
time_ref=time_ref)
gti2 = make_gti({"START": [14], "STOP": [15]}, time_ref=time_ref)
obs1 = SpectrumDatasetOnOff(
counts=on_vector,
counts_off=off_vector1,
aeff=aeff,
edisp=edisp,
livetime=livetime,
mask_safe=np.ones(on_vector.energy.nbin, dtype=bool),
acceptance=1,
acceptance_off=2,
name="1",
gti=gti1,
)
obs2 = SpectrumDatasetOnOff(
counts=on_vector,
counts_off=off_vector2,
aeff=aeff,
edisp=edisp,
livetime=livetime,
mask_safe=np.ones(on_vector.energy.nbin, dtype=bool),
acceptance=1,
acceptance_off=4,
name="2",
gti=gti2,
)
obs_list = [obs1, obs2]
return obs_list
def setup(self):
etrue = np.logspace(-1, 1, 10) * u.TeV
self.e_true = etrue
ereco = np.logspace(-1, 1, 5) * u.TeV
elo = ereco[:-1]
ehi = ereco[1:]
self.e_reco = ereco
self.aeff = EffectiveAreaTable(etrue[:-1], etrue[1:],
np.ones(9) * u.cm**2)
self.edisp = EDispKernel.from_diagonal_response(etrue, ereco)
data = np.ones(elo.shape)
data[-1] = 0 # to test stats calculation with empty bins
self.on_counts = CountsSpectrum(elo, ehi, data)
self.off_counts = CountsSpectrum(elo, ehi, np.ones(elo.shape) * 10)
start = u.Quantity([0], "s")
stop = u.Quantity([1000], "s")
time_ref = Time("2010-01-01 00:00:00.0")
self.gti = GTI.create(start, stop, time_ref)
self.livetime = self.gti.time_sum
self.dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=self.aeff,
edisp=self.edisp,
livetime=self.livetime,
acceptance=np.ones(elo.shape),
acceptance_off=np.ones(elo.shape) * 10,
name="test",
gti=self.gti,
)
def test_fake(self):
"""Test the fake dataset"""
source_model = SkyModel(spectral_model=PowerLawSpectralModel())
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
models=source_model,
aeff=self.aeff,
livetime=self.livetime,
edisp=self.edisp,
acceptance=1,
acceptance_off=10,
)
real_dataset = dataset.copy()
# Define background model counts
elo = self.on_counts.energy.edges[:-1]
ehi = self.on_counts.energy.edges[1:]
data = np.ones(self.on_counts.data.shape)
background_model = CountsSpectrum(elo, ehi, data)
dataset.fake(background_model=background_model,
random_state=314,
name="fake")
assert real_dataset.counts.data.shape == dataset.counts.data.shape
assert real_dataset.counts_off.data.shape == dataset.counts_off.data.shape
assert (real_dataset.counts.energy.center.mean() ==
dataset.counts.energy.center.mean())
assert real_dataset.acceptance.mean() == dataset.acceptance.mean()
assert real_dataset.acceptance_off.mean(
) == dataset.acceptance_off.mean()
assert dataset.counts_off.data.sum() == 39
assert dataset.counts.data.sum() == 5
assert dataset.name == "fake"
def setup(self):
etrue = np.logspace(-1, 1, 10) * u.TeV
self.e_true = etrue
ereco = np.logspace(-1, 1, 5) * u.TeV
elo = ereco[:-1]
ehi = ereco[1:]
self.aeff = EffectiveAreaTable(etrue[:-1], etrue[1:],
np.ones(9) * u.cm**2)
self.edisp = EnergyDispersion.from_diagonal_response(etrue, ereco)
data = np.ones(elo.shape)
data[-1] = 0 # to test stats calculation with empty bins
self.on_counts = CountsSpectrum(elo, ehi, data)
self.off_counts = CountsSpectrum(elo, ehi, np.ones(elo.shape) * 10)
self.livetime = 1000 * u.s
self.dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=self.aeff,
edisp=self.edisp,
livetime=self.livetime,
acceptance=np.ones(elo.shape),
acceptance_off=np.ones(elo.shape) * 10,
obs_id="test",
)
def test_fit_range(self):
"""Test fit range without complication of thresholds"""
dataset = SpectrumDatasetOnOff(counts=self.src,
mask_safe=np.ones(self.src.energy.nbin,
dtype=bool))
dataset.model = self.source_model
assert np.sum(dataset.mask_safe) == self.nbins
e_min, e_max = dataset.energy_range
assert_allclose(e_max.value, 10)
assert_allclose(e_min.value, 0.1)
def test_peek(self):
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=self.aeff,
livetime=self.livetime,
edisp=self.edisp,
acceptance=1,
acceptance_off=10,
)
with mpl_plot_check():
dataset.peek()
def test_str(self):
model = SkyModel(spectral_model=PowerLawSpectralModel())
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
models=model,
aeff=self.aeff,
livetime=self.livetime,
edisp=self.edisp,
acceptance=1,
acceptance_off=10,
)
assert "SpectrumDatasetOnOff" in str(dataset)
assert "wstat" in str(dataset)
def test_plot_fit(self):
model = SkyModel(spectral_model=PowerLawSpectralModel())
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
models=model,
aeff=self.aeff,
livetime=self.livetime,
edisp=self.edisp,
acceptance=1,
acceptance_off=10,
)
with mpl_plot_check():
dataset.plot_fit()
def simulate_spectrum_dataset(model, random_state=0):
energy = np.logspace(-0.5, 1.5, 21) * u.TeV
aeff = EffectiveAreaTable.from_parametrization(energy=energy)
bkg_model = PowerLawSpectralModel(index=2.5, amplitude="1e-12 cm-2 s-1 TeV-1")
dataset = SpectrumDatasetOnOff(
aeff=aeff, model=model, livetime=100 * u.h, acceptance=1, acceptance_off=5
)
eval = SpectrumEvaluator(model=bkg_model, aeff=aeff, livetime=100 * u.h)
bkg_model = eval.compute_npred()
dataset.fake(random_state=random_state, background_model=bkg_model)
return dataset
def test_to_from_ogip_files_no_edisp(self, tmpdir):
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
aeff=self.aeff,
livetime=self.livetime,
mask_safe=np.ones(self.on_counts.energy.nbin, dtype=bool),
acceptance=1,
obs_id="test",
)
dataset.to_ogip_files(outdir=tmpdir, overwrite=True)
filename = tmpdir / "pha_obstest.fits"
newdataset = SpectrumDatasetOnOff.from_ogip_files(filename)
assert_allclose(self.on_counts.data, newdataset.counts.data)
assert newdataset.counts_off is None
assert newdataset.edisp is None
def test_to_from_ogip_files_no_edisp(self, tmp_path):
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
aeff=self.aeff,
livetime=self.livetime,
mask_safe=np.ones(self.on_counts.energy.nbin, dtype=bool),
acceptance=1,
name="test",
)
dataset.to_ogip_files(outdir=tmp_path)
newdataset = SpectrumDatasetOnOff.from_ogip_files(tmp_path /
"pha_obstest.fits")
assert_allclose(self.on_counts.data, newdataset.counts.data)
assert newdataset.counts_off is None
assert newdataset.edisp is None
assert newdataset.gti is None
def test_npred_no_edisp(self):
const = 1 / u.TeV / u.cm**2 / u.s
model = ConstantSpectralModel(const)
livetime = 1 * u.s
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=self.aeff,
model=model,
livetime=livetime,
)
energy = self.aeff.energy.edges * self.aeff.energy.unit
expected = self.aeff.data.data[0] * (energy[-1] -
energy[0]) * const * livetime
assert_allclose(dataset.npred_sig().data.sum(), expected.value)
def test_to_from_ogip_files(self, tmp_path):
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=self.aeff,
edisp=self.edisp,
livetime=self.livetime,
mask_safe=np.ones(self.on_counts.energy.nbin, dtype=bool),
acceptance=1,
acceptance_off=10,
name="test",
gti=self.gti,
)
dataset.to_ogip_files(outdir=tmp_path)
newdataset = SpectrumDatasetOnOff.from_ogip_files(tmp_path /
"pha_obstest.fits")
assert_allclose(self.on_counts.data, newdataset.counts.data)
assert_allclose(self.off_counts.data, newdataset.counts_off.data)
assert_allclose(self.edisp.pdf_matrix, newdataset.edisp.pdf_matrix)
assert_time_allclose(newdataset.gti.time_start, dataset.gti.time_start)
def test_npred_no_edisp(self):
const = 1 * u.Unit("cm-2 s-1 TeV-1")
model = SkyModel(spectral_model=ConstantSpectralModel(const=const))
livetime = 1 * u.s
e_reco = self.on_counts.energy.edges
aeff = EffectiveAreaTable(e_reco[:-1], e_reco[1:],
np.ones(4) * u.cm**2)
dataset = SpectrumDatasetOnOff(
counts=self.on_counts,
counts_off=self.off_counts,
aeff=aeff,
models=model,
livetime=livetime,
)
energy = aeff.energy.edges
expected = aeff.data.data[0] * (energy[-1] -
energy[0]) * const * livetime
assert_allclose(dataset.npred_sig().data.sum(), expected.value)
def test_wstat(self):
"""WStat with on source and background spectrum"""
on_vector = self.src.copy()
on_vector.data += self.bkg.data
obs = SpectrumDatasetOnOff(
counts=on_vector,
counts_off=self.off,
acceptance=1,
acceptance_off=1 / self.alpha,
)
obs.model = self.source_model
self.source_model.parameters.index = 1.12
fit = Fit(obs)
result = fit.run()
pars = self.source_model.parameters
assert_allclose(pars["index"].value, 1.997342, rtol=1e-3)
assert_allclose(pars["amplitude"].value, 100245.187067, rtol=1e-3)
assert_allclose(result.total_stat, 30.022316, rtol=1e-3)
# We assume a PowerLaw shape of the background as well
bkg_model = PowerLaw(index=2.5,
amplitude=1e-11 * u.Unit("cm-2 s-1 TeV-1"),
reference=1 * u.TeV)
evaluator = SpectrumEvaluator(model=bkg_model, aeff=aeff, livetime=livetime)
npred_bkg = evaluator.compute_npred()
# In[ ]:
dataset = SpectrumDatasetOnOff(
aeff=aeff,
edisp=edisp,
model=model_ref,
livetime=livetime,
acceptance=1,
acceptance_off=5,
)
# In[ ]:
get_ipython().run_cell_magic(
'time', '',
'# Now simulate 30 indepenent spectra using the same set of observation conditions.\nn_obs = 100\nseeds = np.arange(n_obs)\n\ndatasets = []\n\nfor idx in range(n_obs):\n dataset.fake(random_state=idx, background_model=npred_bkg)\n datasets.append(dataset.copy())'
)
# Before moving on to the fit let's have a look at the simulated observations.
# In[ ]:
dataset.model = model
dataset.fake(random_state=42)
print(dataset)
# You can see that backgound counts are now simulated
# ### OnOff analysis
#
# To do `OnOff` spectral analysis, which is the usual science case, the standard would be to use `SpectrumDatasetOnOff`, which uses the acceptance to fake off-counts
# In[ ]:
dataset_onoff = SpectrumDatasetOnOff(
aeff=dataset.aeff,
edisp=dataset.edisp,
models=model,
livetime=livetime,
acceptance=1,
acceptance_off=5,
)
dataset_onoff.fake(background_model=dataset.background)
print(dataset_onoff)
# You can see that off counts are now simulated as well. We now simulate several spectra using the same set of observation conditions.
# In[ ]:
get_ipython().run_cell_magic(
'time', '',
'\nn_obs = 100\ndatasets = []\n\nfor idx in range(n_obs):\n dataset_onoff.fake(random_state=idx, background_model=dataset.background)\n dataset_onoff.name = f"obs_{idx}"\n datasets.append(dataset_onoff.copy())'
)