|
1
|
|
|
# Licensed under a 3-clause BSD style license - see LICENSE.rst |
|
2
|
|
|
import pytest |
|
3
|
|
|
from numpy.testing import assert_allclose |
|
4
|
|
|
import astropy.units as u |
|
5
|
|
|
import numpy as np |
|
6
|
|
|
from ...utils.testing import requires_data, requires_dependency |
|
7
|
|
|
from ...utils.random import get_random_state |
|
8
|
|
|
from ...irf import EffectiveAreaTable, EnergyDispersion |
|
9
|
|
|
from ...utils.fitting import Fit |
|
10
|
|
|
from ..models import PowerLaw, ConstantModel, ExponentialCutoffPowerLaw |
|
11
|
|
|
from ...spectrum import PHACountsSpectrum, SpectrumDatasetOnOff |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
class TestSpectrumDatasetOnOff: |
|
15
|
|
|
""" Test ON OFF SpectrumDataset""" |
|
16
|
|
|
|
|
17
|
|
|
def setup(self): |
|
18
|
|
|
|
|
19
|
|
|
etrue = np.logspace(-1, 1, 10) * u.TeV |
|
20
|
|
|
self.e_true = etrue |
|
21
|
|
|
ereco = np.logspace(-1, 1, 5) * u.TeV |
|
22
|
|
|
elo = ereco[:-1] |
|
23
|
|
|
ehi = ereco[1:] |
|
24
|
|
|
|
|
25
|
|
|
self.aeff = EffectiveAreaTable(etrue[:-1], etrue[1:], np.ones(9) * u.cm ** 2) |
|
26
|
|
|
self.edisp = EnergyDispersion.from_diagonal_response(etrue, ereco) |
|
27
|
|
|
|
|
28
|
|
|
self.on_counts = PHACountsSpectrum( |
|
29
|
|
|
elo, ehi, np.ones_like(elo), backscal=np.ones_like(elo) |
|
30
|
|
|
) |
|
31
|
|
|
self.off_counts = PHACountsSpectrum( |
|
32
|
|
|
elo, ehi, np.ones_like(elo) * 10, backscal=np.ones_like(elo) * 10 |
|
33
|
|
|
) |
|
34
|
|
|
|
|
35
|
|
|
self.livetime = 1000 * u.s |
|
36
|
|
|
|
|
37
|
|
|
def test_init_no_model(self): |
|
38
|
|
|
dataset = SpectrumDatasetOnOff( |
|
39
|
|
|
counts_on=self.on_counts, |
|
40
|
|
|
counts_off=self.off_counts, |
|
41
|
|
|
aeff=self.aeff, |
|
42
|
|
|
edisp=self.edisp, |
|
43
|
|
|
livetime=self.livetime, |
|
44
|
|
|
) |
|
45
|
|
|
|
|
46
|
|
|
with pytest.raises(AttributeError): |
|
47
|
|
|
dataset.npred() |
|
48
|
|
|
|
|
49
|
|
|
with pytest.raises(AttributeError): |
|
50
|
|
|
print(dataset.parameters) |
|
51
|
|
|
|
|
52
|
|
|
def test_alpha(self): |
|
53
|
|
|
dataset = SpectrumDatasetOnOff( |
|
54
|
|
|
counts_on=self.on_counts, |
|
55
|
|
|
counts_off=self.off_counts, |
|
56
|
|
|
aeff=self.aeff, |
|
57
|
|
|
edisp=self.edisp, |
|
58
|
|
|
livetime=self.livetime, |
|
59
|
|
|
) |
|
60
|
|
|
|
|
61
|
|
|
assert dataset.alpha.shape == (4,) |
|
62
|
|
|
assert_allclose(dataset.alpha, 0.1) |
|
63
|
|
|
|
|
64
|
|
|
def test_data_shape(self): |
|
65
|
|
|
dataset = SpectrumDatasetOnOff( |
|
66
|
|
|
counts_on=self.on_counts, |
|
67
|
|
|
counts_off=self.off_counts, |
|
68
|
|
|
aeff=self.aeff, |
|
69
|
|
|
edisp=self.edisp, |
|
70
|
|
|
livetime=self.livetime, |
|
71
|
|
|
) |
|
72
|
|
|
|
|
73
|
|
|
assert dataset.data_shape == self.on_counts.data.data.shape |
|
74
|
|
|
|
|
75
|
|
|
def test_npred_no_edisp(self): |
|
76
|
|
|
const = 1 / u.TeV / u.cm ** 2 / u.s |
|
77
|
|
|
model = ConstantModel(const) |
|
78
|
|
|
livetime = 1 * u.s |
|
79
|
|
|
dataset = SpectrumDatasetOnOff( |
|
80
|
|
|
counts_on=self.on_counts, |
|
81
|
|
|
counts_off=self.off_counts, |
|
82
|
|
|
aeff=self.aeff, |
|
83
|
|
|
model=model, |
|
84
|
|
|
livetime=livetime, |
|
85
|
|
|
) |
|
86
|
|
|
|
|
87
|
|
|
expected = ( |
|
88
|
|
|
self.aeff.data.data[0] |
|
89
|
|
|
* (self.aeff.energy.hi[-1] - self.aeff.energy.lo[0]) |
|
90
|
|
|
* const |
|
91
|
|
|
* livetime |
|
92
|
|
|
) |
|
93
|
|
|
|
|
94
|
|
|
assert_allclose(dataset.npred().sum(), expected.value) |
|
95
|
|
|
|
|
96
|
|
|
def test_incorrect_mask(self): |
|
97
|
|
|
mask = np.ones(self.on_counts.data.data.shape, dtype="int") |
|
98
|
|
|
|
|
99
|
|
|
with pytest.raises(ValueError): |
|
100
|
|
|
SpectrumDatasetOnOff( |
|
101
|
|
|
counts_on=self.on_counts, |
|
102
|
|
|
counts_off=self.off_counts, |
|
103
|
|
|
aeff=self.aeff, |
|
104
|
|
|
edisp=self.edisp, |
|
105
|
|
|
livetime=self.livetime, |
|
106
|
|
|
mask=mask, |
|
107
|
|
|
) |
|
108
|
|
|
|
|
109
|
|
|
|
|
110
|
|
|
@requires_dependency("iminuit") |
|
111
|
|
|
class TestSimpleFit: |
|
112
|
|
|
"""Test fit on counts spectra without any IRFs""" |
|
113
|
|
|
|
|
114
|
|
View Code Duplication |
def setup(self): |
|
|
|
|
|
|
115
|
|
|
self.nbins = 30 |
|
116
|
|
|
binning = np.logspace(-1, 1, self.nbins + 1) * u.TeV |
|
117
|
|
|
self.source_model = PowerLaw( |
|
118
|
|
|
index=2, amplitude=1e5 / u.TeV, reference=0.1 * u.TeV |
|
119
|
|
|
) |
|
120
|
|
|
self.bkg_model = PowerLaw(index=3, amplitude=1e4 / u.TeV, reference=0.1 * u.TeV) |
|
121
|
|
|
|
|
122
|
|
|
self.alpha = 0.1 |
|
123
|
|
|
random_state = get_random_state(23) |
|
124
|
|
|
npred = self.source_model.integral(binning[:-1], binning[1:]) |
|
125
|
|
|
source_counts = random_state.poisson(npred) |
|
126
|
|
|
self.src = PHACountsSpectrum( |
|
127
|
|
|
energy_lo=binning[:-1], |
|
128
|
|
|
energy_hi=binning[1:], |
|
129
|
|
|
data=source_counts, |
|
130
|
|
|
backscal=1, |
|
131
|
|
|
) |
|
132
|
|
|
# Currently it's necessary to specify a lifetime |
|
133
|
|
|
self.src.livetime = 1 * u.s |
|
134
|
|
|
|
|
135
|
|
|
npred_bkg = self.bkg_model.integral(binning[:-1], binning[1:]) |
|
136
|
|
|
|
|
137
|
|
|
bkg_counts = random_state.poisson(npred_bkg) |
|
138
|
|
|
off_counts = random_state.poisson(npred_bkg * 1.0 / self.alpha) |
|
139
|
|
|
self.bkg = PHACountsSpectrum( |
|
140
|
|
|
energy_lo=binning[:-1], energy_hi=binning[1:], data=bkg_counts |
|
141
|
|
|
) |
|
142
|
|
|
self.off = PHACountsSpectrum( |
|
143
|
|
|
energy_lo=binning[:-1], |
|
144
|
|
|
energy_hi=binning[1:], |
|
145
|
|
|
data=off_counts, |
|
146
|
|
|
backscal=1.0 / self.alpha, |
|
147
|
|
|
) |
|
148
|
|
|
|
|
149
|
|
|
def test_wstat(self): |
|
150
|
|
|
"""WStat with on source and background spectrum""" |
|
151
|
|
|
on_vector = self.src.copy() |
|
152
|
|
|
on_vector.data.data += self.bkg.data.data |
|
153
|
|
|
obs = SpectrumDatasetOnOff(counts_on=on_vector, counts_off=self.off) |
|
154
|
|
|
obs.model = self.source_model |
|
155
|
|
|
|
|
156
|
|
|
self.source_model.parameters.index = 1.12 |
|
157
|
|
|
|
|
158
|
|
|
fit = Fit(obs) |
|
159
|
|
|
result = fit.run() |
|
160
|
|
|
pars = self.source_model.parameters |
|
161
|
|
|
|
|
162
|
|
|
assert_allclose(pars["index"].value, 1.997342, rtol=1e-3) |
|
163
|
|
|
assert_allclose(pars["amplitude"].value, 100245.187067, rtol=1e-3) |
|
164
|
|
|
assert_allclose(result.total_stat, 30.022316, rtol=1e-3) |
|
165
|
|
|
|
|
166
|
|
|
def test_joint(self): |
|
167
|
|
|
"""Test joint fit for obs with different energy binning""" |
|
168
|
|
|
on_vector = self.src.copy() |
|
169
|
|
|
on_vector.data.data += self.bkg.data.data |
|
170
|
|
|
obs1 = SpectrumDatasetOnOff(counts_on=on_vector, counts_off=self.off) |
|
171
|
|
|
obs1.model = self.source_model |
|
172
|
|
|
|
|
173
|
|
|
src_rebinned = self.src.rebin(2) |
|
174
|
|
|
bkg_rebinned = self.off.rebin(2) |
|
175
|
|
|
src_rebinned.data.data += self.bkg.rebin(2).data.data |
|
176
|
|
|
|
|
177
|
|
|
obs2 = SpectrumDatasetOnOff(counts_on=src_rebinned, counts_off=bkg_rebinned) |
|
178
|
|
|
obs2.model = self.source_model |
|
179
|
|
|
|
|
180
|
|
|
fit = Fit([obs1, obs2]) |
|
181
|
|
|
fit.run() |
|
182
|
|
|
pars = self.source_model.parameters |
|
183
|
|
|
assert_allclose(pars["index"].value, 1.996456, rtol=1e-3) |
|
184
|
|
|
|
|
185
|
|
|
|
|
186
|
|
|
@requires_data("gammapy-data") |
|
187
|
|
|
@requires_dependency("iminuit") |
|
188
|
|
|
class TestSpectralFit: |
|
189
|
|
|
"""Test fit in astrophysical scenario""" |
|
190
|
|
|
|
|
191
|
|
|
def setup(self): |
|
192
|
|
|
path = "$GAMMAPY_DATA/joint-crab/spectra/hess/" |
|
193
|
|
|
obs1 = SpectrumDatasetOnOff.read(path + "pha_obs23523.fits") |
|
194
|
|
|
obs2 = SpectrumDatasetOnOff.read(path + "pha_obs23592.fits") |
|
195
|
|
|
self.obs_list = [obs1, obs2] |
|
196
|
|
|
|
|
197
|
|
|
self.pwl = PowerLaw( |
|
198
|
|
|
index=2, amplitude=1e-12 * u.Unit("cm-2 s-1 TeV-1"), reference=1 * u.TeV |
|
199
|
|
|
) |
|
200
|
|
|
|
|
201
|
|
|
self.ecpl = ExponentialCutoffPowerLaw( |
|
202
|
|
|
index=2, |
|
203
|
|
|
amplitude=1e-12 * u.Unit("cm-2 s-1 TeV-1"), |
|
204
|
|
|
reference=1 * u.TeV, |
|
205
|
|
|
lambda_=0.1 / u.TeV, |
|
206
|
|
|
) |
|
207
|
|
|
|
|
208
|
|
|
# Example fit for one observation |
|
209
|
|
|
self.obs_list[0].model = self.pwl |
|
210
|
|
|
self.fit = Fit(self.obs_list[0]) |
|
211
|
|
|
|
|
212
|
|
|
def set_model(self, model): |
|
213
|
|
|
for obs in self.obs_list: |
|
214
|
|
|
obs.model = model |
|
215
|
|
|
|
|
216
|
|
|
@requires_dependency("iminuit") |
|
217
|
|
|
def test_basic_results(self): |
|
218
|
|
|
self.set_model(self.pwl) |
|
219
|
|
|
result = self.fit.run() |
|
220
|
|
|
pars = self.fit.datasets.parameters |
|
221
|
|
|
|
|
222
|
|
|
assert self.pwl is self.obs_list[0].model |
|
223
|
|
|
|
|
224
|
|
|
assert_allclose(result.total_stat, 38.343, rtol=1e-3) |
|
225
|
|
|
assert_allclose(pars["index"].value, 2.817, rtol=1e-3) |
|
226
|
|
|
assert pars["amplitude"].unit == "cm-2 s-1 TeV-1" |
|
227
|
|
|
assert_allclose(pars["amplitude"].value, 5.142e-11, rtol=1e-3) |
|
228
|
|
|
assert_allclose(self.obs_list[0].npred()[60], 0.6102, rtol=1e-3) |
|
229
|
|
|
pars.to_table() |
|
230
|
|
|
|
|
231
|
|
|
def test_basic_errors(self): |
|
232
|
|
|
self.set_model(self.pwl) |
|
233
|
|
|
self.fit.run() |
|
234
|
|
|
pars = self.fit.datasets.parameters |
|
235
|
|
|
|
|
236
|
|
|
assert_allclose(pars.error("index"), 0.1496, rtol=1e-3) |
|
237
|
|
|
assert_allclose(pars.error("amplitude"), 6.423e-12, rtol=1e-3) |
|
238
|
|
|
pars.to_table() |
|
239
|
|
|
|
|
240
|
|
|
def test_compound(self): |
|
241
|
|
|
model = self.pwl * 2 |
|
242
|
|
|
self.set_model(model) |
|
243
|
|
|
fit = Fit(self.obs_list[0]) |
|
244
|
|
|
fit.run() |
|
245
|
|
|
pars = fit.datasets.parameters |
|
246
|
|
|
|
|
247
|
|
|
assert_allclose(pars["index"].value, 2.8166, rtol=1e-3) |
|
248
|
|
|
p = pars["amplitude"] |
|
249
|
|
|
assert p.unit == "cm-2 s-1 TeV-1" |
|
250
|
|
|
assert_allclose(p.value, 5.0714e-12, rtol=1e-3) |
|
251
|
|
|
|
|
252
|
|
|
def test_ecpl_fit(self): |
|
253
|
|
|
self.set_model(self.ecpl) |
|
254
|
|
|
fit = Fit(self.obs_list[0]) |
|
255
|
|
|
fit.run() |
|
256
|
|
|
|
|
257
|
|
|
actual = fit.datasets.parameters["lambda_"].quantity |
|
258
|
|
|
assert actual.unit == "TeV-1" |
|
259
|
|
|
assert_allclose(actual.value, 0.145215, rtol=1e-2) |
|
260
|
|
|
|
|
261
|
|
|
def test_joint_fit(self): |
|
262
|
|
|
self.set_model(self.pwl) |
|
263
|
|
|
fit = Fit(self.obs_list) |
|
264
|
|
|
fit.run() |
|
265
|
|
|
actual = fit.datasets.parameters["index"].value |
|
266
|
|
|
assert_allclose(actual, 2.7806, rtol=1e-3) |
|
267
|
|
|
|
|
268
|
|
|
actual = fit.datasets.parameters["amplitude"].quantity |
|
269
|
|
|
assert actual.unit == "cm-2 s-1 TeV-1" |
|
270
|
|
|
assert_allclose(actual.value, 5.200e-11, rtol=1e-3) |
|
271
|
|
|
|
gammapy /
gammapy
Loading history...