Coverage for lst_auto_rta/Spectra.py: 0%
193 statements
« prev ^ index » next coverage.py v7.6.1, created at 2024-09-18 19:29 +0000
1#!/usr/bin/env python
3import astropy
4import gammapy
5import matplotlib
6import numpy as np
7import regions
9print("gammapy:", gammapy.__version__)
10print("numpy:", np.__version__)
11print("astropy", astropy.__version__)
12print("regions", regions.__version__)
13print("matplotlib", matplotlib.__version__)
15import os
17import astropy.units as u
18import matplotlib.pyplot as plt
19import matplotlib.style as style
20import numpy as np
21import scipy
22from astropy.coordinates import SkyCoord
24style.use("tableau-colorblind10")
25import argparse
26import pickle
27from pathlib import Path
29from gammapy.data import DataStore
30from gammapy.datasets import Datasets, FluxPointsDataset, MapDataset, SpectrumDataset
31from gammapy.estimators import FluxPointsEstimator
32from gammapy.makers import (
33 MapDatasetMaker,
34 ReflectedRegionsBackgroundMaker,
35 SafeMaskMaker,
36 SpectrumDatasetMaker,
37)
38from gammapy.makers.utils import make_theta_squared_table
39from gammapy.maps import Map, MapAxis, RegionGeom, WcsGeom
40from gammapy.modeling import Fit
41from gammapy.modeling.models import (
42 ExpCutoffPowerLawSpectralModel,
43 LogParabolaSpectralModel,
44 PowerLawSpectralModel,
45 SkyModel,
46)
47from gammapy.visualization import plot_spectrum_datasets_off_regions, plot_theta_squared_table
48from matplotlib.offsetbox import AnchoredText
49from regions import CircleSkyRegion
51parser = argparse.ArgumentParser(
52 description="Automatic Script for the DL1 check", formatter_class=argparse.ArgumentDefaultsHelpFormatter
53)
54parser.add_argument("-d", "--directory", default="/fefs/onsite/pipeline/rta/data/", help="Directory for data")
55parser.add_argument("-da", "--date", default="20230705", help="Date of the run to check")
56parser.add_argument("-r", "--run-id", default="13600", help="run id to check")
57parser.add_argument("-add", "--add-string", default="", help="add a string to the path")
58parser.add_argument("-RA", "--right-ascension", default="270.19042", help="right-ascension in deg")
59parser.add_argument("-DEC", "--declination", default="78.46806", help="declination in deg")
61args = parser.parse_args()
62config = vars(args)
64print(config)
66location_data = (
67 config["directory"] + config["date"] + "/" + config["run_id"] + "/" + config["add_string"] + "/DL3"
68) # path to DL3 folder
69source_name = config["run_id"] # e.g., Crab, GRB210807A
70cut_type = "standard" # e.g., loose, hard, ...
71filename_output = "{name}_{cut}".format(name=source_name, cut=cut_type)
74source_position = SkyCoord(ra=config["right_ascension"], dec=config["declination"], unit="deg", frame="icrs")
75max_offset_run = 5 * u.deg
76work_directory = location_data
77path_plot = Path(work_directory + "/../plots")
78print(work_directory + "/../plots")
79path_plot.mkdir(exist_ok=True)
80path_background = Path(work_directory + "/../plots")
81path_background.mkdir(exist_ok=True)
83on_radius = 0.2 * u.deg
84exclusion_radius = 0.35 * u.deg
86e_min_reco = 0.02 * u.TeV
87e_max_reco = 10.0 * u.TeV
88n_bin_per_decade_e_reco = 10
90e_min_true = 0.005 * u.TeV
91e_max_true = 40.0 * u.TeV
92n_bin_per_decade_e_true = 20
94e_min_flux_point = 0.02 * u.TeV
95e_max_flux_point = 10.0 * u.TeV
96n_bin_per_decade_flux_point = 5
99def plot_excess(dataset_to_plot, ax=None, kwargs_excess=None, kwargs_npred_signal=None, **kwargs):
100 from gammapy.stats import CashCountsStatistic, WStatCountsStatistic
102 region = None
104 kwargs_excess = kwargs_excess or {}
105 kwargs_npred_signal = kwargs_npred_signal or {}
107 counts, npred = dataset_to_plot.counts.copy(), dataset_to_plot.npred()
109 if dataset_to_plot.mask is None:
110 mask = dataset_to_plot.copy()
111 mask.data = 1
112 else:
113 mask = dataset_to_plot.mask
114 counts *= mask
115 npred *= mask
117 if dataset_to_plot.stat_type == "wstat":
118 counts_off = dataset_to_plot.counts_off * mask
119 with np.errstate(invalid="ignore"):
120 alpha = dataset_to_plot.alpha * mask
122 stat = WStatCountsStatistic(
123 n_on=counts,
124 n_off=counts_off,
125 alpha=alpha,
126 # mu_sig=npred,
127 )
128 elif dataset_to_plot.stat_type == "cash":
129 stat = CashCountsStatistic(counts.data, npred.data)
131 yerr = np.zeros((2,) + counts.data.shape)
132 yerr[0], yerr[1] = -stat.compute_errn(), stat.compute_errp()
133 yerr = stat.error
135 plot_kwargs = kwargs.copy()
136 plot_kwargs.update(kwargs_excess)
137 plot_kwargs.setdefault("label", "Excess counts")
138 ax = dataset_to_plot.excess.plot(ax, yerr=yerr, **plot_kwargs)
140 plot_kwargs = kwargs.copy()
141 plot_kwargs.update(kwargs_npred_signal)
142 plot_kwargs.setdefault("label", "Predicted signal counts")
143 dataset_to_plot.npred_signal().plot_hist(ax, **plot_kwargs)
145 ax.legend(numpoints=1)
146 return ax
149def plot_fit(
150 dataset_to_plot,
151 ax_spectrum=None,
152 ax_residuals=None,
153 kwargs_spectrum=None,
154 kwargs_residuals=None,
155):
156 from gammapy.datasets.utils import get_axes
157 from matplotlib.gridspec import GridSpec
159 gs = GridSpec(7, 1)
160 ax_spectrum, ax_residuals = get_axes(
161 ax_spectrum,
162 ax_residuals,
163 8,
164 7,
165 [gs[:5, :]],
166 [gs[5:, :]],
167 kwargs2={"sharex": ax_spectrum},
168 )
169 kwargs_spectrum = kwargs_spectrum or {}
170 kwargs_residuals = kwargs_residuals or {}
172 plot_excess(dataset_to_plot, ax_spectrum, **kwargs_spectrum)
174 dataset_to_plot.plot_residuals_spectral(ax_residuals, **kwargs_residuals)
176 method = kwargs_residuals.get("method", "diff")
177 label = dataset_to_plot._residuals_labels[method]
178 ax_residuals.set_ylabel(f"Residuals\n{label}")
180 return ax_spectrum, ax_residuals
183data_store = DataStore.from_dir(location_data)
184obs_ids = data_store.obs_table[source_position.separation(data_store.obs_table.pointing_radec) < max_offset_run][
185 "OBS_ID"
186]
187obs_collection = data_store.get_observations(obs_ids, required_irf="point-like")
189on_region = CircleSkyRegion(center=source_position, radius=on_radius)
190exclude_region = CircleSkyRegion(center=source_position, radius=exclusion_radius)
192n_bin_energy_reco = int(
193 (np.log10(e_max_reco.to_value(u.TeV)) - np.log10(e_min_reco.to_value(u.TeV))) * n_bin_per_decade_e_reco
194)
195energy_axis_reco = MapAxis.from_edges(
196 np.logspace(np.log10(e_min_reco.to_value(u.TeV)), np.log10(e_max_reco.to_value(u.TeV)), n_bin_energy_reco),
197 unit="TeV",
198 name="energy",
199 interp="log",
200)
201n_bin_energy_true = int(
202 (np.log10(e_max_true.to_value(u.TeV)) - np.log10(e_min_true.to_value(u.TeV))) * n_bin_per_decade_e_true
203)
204energy_axis_true = MapAxis.from_edges(
205 np.logspace(np.log10(e_min_true.to_value(u.TeV)), np.log10(e_max_true.to_value(u.TeV)), n_bin_energy_true),
206 unit="TeV",
207 name="energy_true",
208 interp="log",
209)
211geom_on_region = RegionGeom.create(region=on_region, axes=[energy_axis_reco])
212geom = WcsGeom.create(
213 skydir=source_position,
214 npix=(200, 200),
215 binsz=0.05,
216 frame="icrs",
217 axes=[energy_axis_reco],
218)
220geom_image = geom.to_image()
221exclusion_mask = ~geom_image.region_mask([exclude_region])
224dataset_maker_spectrum = SpectrumDatasetMaker(selection=["counts", "exposure", "edisp"], use_region_center=True)
225spectrum_dataset_empty = SpectrumDataset.create(geom=geom_on_region, energy_axis_true=energy_axis_true)
226bkg_maker_spectrum = ReflectedRegionsBackgroundMaker(exclusion_mask=exclusion_mask)
228dataset_stacked_map = MapDataset.create(geom=geom, name=source_name + "_stacked")
229map_dataset_maker = MapDatasetMaker(selection=["counts"])
230map_maker_safe_mask = SafeMaskMaker(methods=["offset-max"], offset_max="4 deg")
232datasets_spectrum_joint = Datasets()
233for obs in obs_collection:
234 dataset_spectrum = dataset_maker_spectrum.run(spectrum_dataset_empty.copy(name=f"obs-{obs.obs_id}"), obs)
235 dataset_on_off_spectrum = bkg_maker_spectrum.run(observation=obs, dataset=dataset_spectrum)
236 datasets_spectrum_joint.append(dataset_on_off_spectrum)
238 map_cutout = dataset_stacked_map.cutout(obs.pointing_radec, width="6.5 deg")
239 map_dataset = map_dataset_maker.run(map_cutout, obs)
240 map_dataset = map_maker_safe_mask.run(map_dataset, obs)
241 dataset_stacked_map.stack(map_dataset)
243info_table = datasets_spectrum_joint.info_table(cumulative=True)
245reference_energy = energy_axis_reco.center[int(energy_axis_reco.nbin / 2)]
247spectral_model = {
248 "power_law": PowerLawSpectralModel(
249 amplitude=1e-12 * u.Unit("cm-2 s-1 TeV-1"), index=2, reference=reference_energy
250 ),
251 "log_parabola": LogParabolaSpectralModel(
252 amplitude=1e-12 * u.Unit("cm-2 s-1 TeV-1"), alpha=2, beta=0, reference=reference_energy
253 ),
254 "power_law_exponantial_cutoff": ExpCutoffPowerLawSpectralModel(
255 amplitude=1e-12 * u.Unit("cm-2 s-1 TeV-1"), index=2, lambda_=1.0 / reference_energy, reference=reference_energy
256 ),
257}
258model = {}
259for k in spectral_model.keys():
260 model[k] = SkyModel(spectral_model=spectral_model[k], name=source_name + "_" + k)
262datasets_spectrum_joint_with_model = {}
263datasets_spectrum_joint_with_model_stacked = {}
264fit_joint = {}
265result_fit_joint = {}
266for k in spectral_model.keys():
267 datasets_spectrum_joint_with_model[k] = datasets_spectrum_joint.copy()
268 datasets_spectrum_joint_with_model[k].models = [model[k]]
269 fit_joint[k] = Fit()
270 result_fit_joint[k] = fit_joint[k].run(datasets=datasets_spectrum_joint_with_model[k])
272 datasets_spectrum_joint_with_model_stacked[k] = datasets_spectrum_joint_with_model[k].stack_reduce()
273 datasets_spectrum_joint_with_model_stacked[k].models = [model[k]]
275if result_fit_joint["power_law_exponantial_cutoff"].total_stat > result_fit_joint["log_parabola"].total_stat:
276 chalenger_model = "log_parabola"
277 print("Log parabola perform better than power law than exponantial cut off")
278else:
279 chalenger_model = "power_law_exponantial_cutoff"
280 print("Power law than exponantial cut off perform better than log parabola")
283# significance_chalenger = np.sqrt((result_fit_joint['power_law'].total_stat-result_fit_joint[chalenger_model].total_stat))
284print("\n\nLikelihood Ratio test\n")
285LR_statistic = result_fit_joint["power_law"].total_stat - result_fit_joint[chalenger_model].total_stat
286dof = 1
287p_val = scipy.stats.chi2.sf(LR_statistic, dof)
289significance_chalenger = np.sqrt(2) * scipy.special.erfinv(1 - p_val)
290print(
291 "The significance of the {} model over power law is {:.5f} sigma".format(chalenger_model, significance_chalenger)
292)
294if significance_chalenger > 5.0:
295 selected_model = chalenger_model
296 print("The spectral model " + chalenger_model + " is selected as significant")
297else:
298 selected_model = "power_law"
299 print("The spectral model " + chalenger_model + " is rejected as not significant over power law")
300 print("The selected spectral model is the power law")
302ax1, ax2 = plot_fit(datasets_spectrum_joint_with_model_stacked[k])
303excess = datasets_spectrum_joint_with_model_stacked[k].excess.data
304n_on = datasets_spectrum_joint_with_model_stacked[k].counts.data
305n_off = datasets_spectrum_joint_with_model_stacked[k].counts_off.data
306alpha = datasets_spectrum_joint_with_model_stacked[k].alpha.data
307low_excess = excess - np.sqrt(n_on + alpha * alpha * n_off)
308low_excess = low_excess[low_excess > 0.0]
309ax1.set_ylim(bottom=np.min(0.8 * low_excess))
311print(datasets_spectrum_joint_with_model[selected_model].models)
313n_bin_energy_flux_point = int(
314 (np.log10(e_max_flux_point.to_value(u.TeV)) - np.log10(e_min_flux_point.to_value(u.TeV)))
315 * n_bin_per_decade_flux_point
316)
317energy_edges_flux_point = (
318 np.logspace(
319 np.log10(e_min_flux_point.to_value(u.TeV)), np.log10(e_max_flux_point.to_value(u.TeV)), n_bin_energy_flux_point
320 )
321 * u.TeV
322)
324fpe = FluxPointsEstimator(
325 energy_edges=energy_edges_flux_point,
326 source=source_name + "_" + selected_model,
327 selection_optional="all",
328 n_sigma_ul=3,
329 reoptimize=False,
330)
331flux_points = fpe.run(datasets=datasets_spectrum_joint_with_model[selected_model])
333flux_points.to_table(sed_type="e2dnde", formatted=True)
335plt.figure(figsize=(8, 5))
336ax = flux_points.plot(sed_type="e2dnde", color="darkorange")
337flux_points.plot_ts_profiles(ax=ax, sed_type="e2dnde")
338plt.savefig(os.path.join(path_plot, "{}__flux_point.png".format(filename_output)), dpi=300)
340flux_points_dataset = FluxPointsDataset(data=flux_points, models=model[selected_model])
341flux_points_dataset.plot_fit()
342plt.savefig(os.path.join(path_plot, "{}__spectra.png".format(filename_output)), dpi=300)