def get_spectrum(filename, data_id="1"):
ui.load_data(id=data_id, filename=filename)
d = ui.get_data(data_id)
arf = d.get_arf()
rmf = d.get_rmf()
return d, arf, rmf
def load_data(ui, make_data_path):
"""
Load dataset before every test.
"""
ui.load_data(make_data_path("3c273.pi"))
ui.set_source("powlaw1d.p")
ui.set_bkg_model("const1d.c")
ui.fit()
def load_data(ui, make_data_path):
"""
Load dataset before every test.
"""
ui.load_data(make_data_path("3c273.pi"))
ui.set_source("powlaw1d.p")
ui.set_bkg_model("const1d.c")
ui.fit()
def fit_sherpa():
sh.load_data(obs_path)
sh.set_source("powlaw1d.model")
model.ref = 1e9 # 1 TeV = 1e9 keV
model.ampl = 1.23e-20 # in cm**-2 s**-1 keV**-1
model.gamma = 2.8
sh.set_stat("cash")
sh.notice(energy_range[0].to("keV").value, energy_range[1].to("keV").value)
sh.fit()
print(sh.get_fit_results())
def setup(imgfile, emapfile):
shp.load_data(imgfile)
shp.load_table_model('emap', emapfile)
shp.freeze(emap.ampl)
# we're using an on axis psf provided in
# /packages/CALDB/data/nustar/fpm/bcf/psf/
shp.load_psf('psf', 'nuA2dpsf20100101v003_onaxis.fits')
shp.set_psf(psf)
psf.center = (500.0, 500.0)
psf.size = (200, 200)
def run_hspec_fit(self, model, thres_low, thres_high):
"""Run the gammapy.hspec fit
Parameters
----------
model : str
Sherpa model
thres_high : `~gammapy.spectrum.Energy`
Upper threshold of the spectral fit
thres_low : `~gammapy.spectrum.Energy`
Lower threshold of the spectral fit
"""
log.info("Starting HSPEC")
import sherpa.astro.ui as sau
from ..hspec import wstat
from sherpa.models import PowLaw1D
if model == 'PL':
p1 = PowLaw1D('p1')
p1.gamma = 2.2
p1.ref = 1e9
p1.ampl = 6e-19
else:
raise ValueError('Desired Model is not defined')
thres = thres_low.to('keV').value
emax = thres_high.to('keV').value
sau.freeze(p1.ref)
sau.set_conf_opt("max_rstat", 100)
list_data = []
for obs in self.observations:
datid = obs.phafile.parts[-1][7:12]
sau.load_data(datid, str(obs.phafile))
sau.notice_id(datid, thres, emax)
sau.set_source(datid, p1)
list_data.append(datid)
wstat.wfit(list_data)
sau.covar()
fit_val = sau.get_covar_results()
fit_attrs = ('parnames', 'parvals', 'parmins', 'parmaxes')
fit = dict((attr, getattr(fit_val, attr)) for attr in fit_attrs)
fit = self.apply_containment(fit)
sau.clean()
self.fit = fit
def image_model_sherpa(exposure,
psf,
sources,
model_image,
overwrite):
"""Compute source model image with Sherpa.
Inputs:
* Source list (JSON file)
* PSF (JSON file)
* Exposure image (FITS file)
Outputs:
* Source model flux image (FITS file)
* Source model excess image (FITS file)
"""
import sherpa.astro.ui as sau
from ..image.models.psf import Sherpa
from ..image.models.utils import read_json
log.info('Reading exposure: {0}'.format(exposure))
# Note: We don't really need the exposure as data,
# but this is a simple way to init the dataspace to the correct shape
sau.load_data(exposure)
sau.load_table_model('exposure', exposure)
log.info('Reading PSF: {0}'.format(psf))
Sherpa(psf).set()
log.info('Reading sources: {0}'.format(sources))
read_json(sources, sau.set_source)
name = sau.get_source().name
full_model = 'exposure * psf({})'.format(name)
sau.set_full_model(full_model)
log.info('Computing and writing model_image: {0}'.format(model_image))
sau.save_model(model_image, clobber=overwrite)
sau.clean()
sau.delete_psf()
def test_eqwith_err(make_data_path, restore_xspec_settings):
def check(a0, a1, a2):
assert a0 == approx(0.16443033244310976, rel=1e-3)
assert a1 == approx(0.09205564216156815, rel=1e-3)
assert a2 == approx(0.23933118287470895, rel=1e-3)
ui.set_method('neldermead')
ui.set_stat('cstat')
ui.set_xsabund('angr')
ui.set_xsxsect('bcmc')
ui.load_data(make_data_path('12845.pi'))
ui.notice(0.5, 7)
ui.set_model("xsphabs.gal*xszphabs.zabs*(powlaw1d.p1+xszgauss.g1)")
ui.set_par(gal.nh, 0.08)
ui.freeze(gal)
ui.set_par(zabs.redshift, 0.518)
ui.set_par(g1.redshift, 0.518)
ui.set_par(g1.Sigma, 0.01)
ui.freeze(g1.Sigma)
ui.set_par(g1.LineE, min=6.0, max=7.0)
ui.fit()
numpy.random.seed(12345)
result = ui.eqwidth(p1, p1 + g1, error=True, niter=100)
check(result[0], result[1], result[2])
params = result[3]
numpy.random.seed(12345)
result = ui.eqwidth(p1, p1 + g1, error=True, params=params, niter=100)
check(result[0], result[1], result[2])
parvals = ui.get_fit_results().parvals
assert parvals[0] == approx(0.6111340686157877, rel=1.0e-3)
assert parvals[1] == approx(1.6409785803466297, rel=1.0e-3)
assert parvals[2] == approx(8.960926761312153e-05, rel=1.0e-3)
assert parvals[3] == approx(6.620017726014523, rel=1.0e-3)
assert parvals[4] == approx(1.9279114810359657e-06, rel=1.0e-3)
def test_eqwith_err(make_data_path, restore_xspec_settings):
def check(a0, a1, a2):
assert a0 == pytest.approx(0.16443033244310976, rel=1e-3)
assert a1 == pytest.approx(0.09205564216156815, rel=1e-3)
assert a2 == pytest.approx(0.23933118287470895, rel=1e-3)
ui.set_method('neldermead')
ui.set_stat('cstat')
ui.set_xsabund('angr')
ui.set_xsxsect('bcmc')
ui.load_data(make_data_path('12845.pi'))
ui.notice(0.5, 7)
ui.set_model("xsphabs.gal*xszphabs.zabs*(powlaw1d.p1+xszgauss.g1)")
ui.set_par(gal.nh, 0.08)
ui.freeze(gal)
ui.set_par(zabs.redshift, 0.518)
ui.set_par(g1.redshift, 0.518)
ui.set_par(g1.Sigma, 0.01)
ui.freeze(g1.Sigma)
ui.set_par(g1.LineE, min=6.0, max=7.0)
ui.fit()
np.random.seed(12345)
result = ui.eqwidth(p1, p1 + g1, error=True, niter=100)
check(result[0], result[1], result[2])
params = result[3]
np.random.seed(12345)
result = ui.eqwidth(p1, p1 + g1, error=True, params=params, niter=100)
check(result[0], result[1], result[2])
parvals = ui.get_fit_results().parvals
assert parvals[0] == pytest.approx(0.6111340686157877, rel=1.0e-3)
assert parvals[1] == pytest.approx(1.6409785803466297, rel=1.0e-3)
assert parvals[2] == pytest.approx(8.960926761312153e-05, rel=1.0e-3)
assert parvals[3] == pytest.approx(6.620017726014523, rel=1.0e-3)
assert parvals[4] == pytest.approx(1.9279114810359657e-06, rel=1.0e-3)
def sherpa_model_image(exposure, psf, sources, model_image, overwrite):
"""Compute source model image with Sherpa.
Inputs:
* Source list (JSON file)
* PSF (JSON file)
* Exposure image (FITS file)
Outputs:
* Source model flux image (FITS file)
* Source model excess image (FITS file)
"""
import logging
logging.basicConfig(level=logging.DEBUG,
format='%(levelname)s - %(message)s')
import sherpa.astro.ui as sau # @UnresolvedImport
from ..morphology.psf import Sherpa
from ..morphology.utils import read_json
logging.info('Reading exposure: {0}'.format(exposure))
# Note: We don't really need the exposure as data,
# but this is a simple way to init the dataspace to the correct shape
sau.load_data(exposure)
sau.load_table_model('exposure', exposure)
logging.info('Reading PSF: {0}'.format(psf))
Sherpa(psf).set()
logging.info('Reading sources: {0}'.format(sources))
read_json(sources, sau.set_source)
name = sau.get_source().name
full_model = 'exposure * psf({})'.format(name)
sau.set_full_model(full_model)
logging.info('Computing and writing model_image: {0}'.format(model_image))
sau.save_model(model_image, clobber=overwrite)
def _run_hspec_fit(self):
"""Run the gammapy.hspec fit
"""
log.info("Starting HSPEC")
import sherpa.astro.ui as sau
from ..hspec import wstat
sau.set_conf_opt("max_rstat", 100)
thres_lo = self.energy_threshold_low.to('keV').value
thres_hi = self.energy_threshold_high.to('keV').value
sau.freeze(self.model.ref)
list_data = []
for pha in self.pha:
datid = pha.parts[-1][7:12]
sau.load_data(datid, str(pha))
sau.notice_id(datid, thres_lo, thres_hi)
sau.set_source(datid, self.model)
list_data.append(datid)
wstat.wfit(list_data)
# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Compute results with Sherpa"""
import numpy as np
import sherpa.astro.ui as sau
sau.load_data('counts.fits.gz')
sau.set_source('normgauss2d.source + const2d.background')
sau.set_stat('cstat')
# Ask for high-precision results
sau.set_method_opt('ftol', 1e-20)
sau.set_covar_opt('eps', 1e-20)
# Set start parameters close to simulation values to make the fit converge
sau.set_par('source.xpos', 101)
sau.set_par('source.ypos', 101)
sau.set_par('source.ampl', 1.1e3)
sau.set_par('source.fwhm', 10)
sau.set_par('background.c0', 1.1)
# Run fit and covariance estimation
# Results are automatically printed to the screen
sau.fit()
sau.covar()
# Sherpa uses fwhm instead of sigma as extension parameter ... need to convert
# http://cxc.harvard.edu/sherpa/ahelp/gauss2d.html
fwhm_to_sigma = 1. / np.sqrt(8 * np.log(2))
cov = sau.get_covar_results()
sigma = fwhm_to_sigma * cov.parvals[0]
sigma_err = fwhm_to_sigma * cov.parmaxes[0]
def test_save_image_dataimg_fits_wcs(writer, make_data_path, tmp_path):
"""Does save_image work for a FITS image with WCS
We also check the input file, is read in, for fun.
dmlist reports that this file has
Block 1: EVENTS_IMAGE Image Int2(316x313)
Physical Axis Transforms for Image Block EVENTS_IMAGE
Group# Axis#
1 1,2 sky(x) = (+2986.8401) +(+1.0)* ((#1)-(+0.50))
(y) (+4362.6602) (+1.0) ((#2) (+0.50))
World Coordinate Axis Transforms for Image Block EVENTS_IMAGE
Group# Axis#
1 1,2 EQPOS(RA ) = (+149.8853)[deg] +TAN[(-0.000136667)* (sky(x)-(+4096.50))]
(DEC) (+2.6079 ) (+0.000136667) ( (y) (+4096.50))
"""
from sherpa.astro.io import read_image
from sherpa.astro.io.meta import Meta
from sherpa.astro.io.wcs import WCS
# It looks like we don't write out the header info, at least for
# pyfits. Probably a bug.
#
def check_data(d, header=True):
assert isinstance(d, ui.DataIMG)
assert d.shape == (313, 316)
assert d.y.shape == (313 * 316, )
assert d.y.sum() == 965
assert d.y.max() == 3
hdr = d.header
assert isinstance(hdr, Meta)
if header:
assert hdr['OBJECT'] == 'CSC'
assert hdr['ONTIME'] == pytest.approx(18220.799932122)
assert hdr['TSTOP'] == pytest.approx(280770182.77747)
assert isinstance(d.sky, WCS)
assert d.sky.name == 'physical'
assert d.sky.type == 'LINEAR'
assert d.sky.crval == pytest.approx([2986.84008789, 4362.66015625])
assert d.sky.crpix == pytest.approx([0.5, 0.5])
assert d.sky.cdelt == pytest.approx([1, 1])
assert isinstance(d.eqpos, WCS)
assert d.eqpos.name == 'world'
assert d.eqpos.type == 'WCS'
assert d.eqpos.epoch == pytest.approx(2000)
assert d.eqpos.equinox == pytest.approx(2000)
assert d.eqpos.crval == pytest.approx([149.88533198, 2.60794887])
assert d.eqpos.crpix == pytest.approx([4096.5, 4096.5])
assert d.eqpos.cdelt * 3600 == pytest.approx([-0.492, 0.492])
infile = make_data_path('acisf08478_000N001_r0043_regevt3_srcimg.fits')
ui.load_data(2, infile)
dorig = ui.get_data(2)
check_data(dorig)
out = tmp_path / "data.dat"
outfile = str(out)
writer(2, outfile, ascii=False)
ans = read_image(outfile)
check_data(ans, header=False)
# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Compute results with Sherpa"""
from __future__ import print_function, division
# __doctest_skip__
__doctest_skip__ = ['*']
import numpy as np
import sherpa.astro.ui as sau
sau.load_data('counts.fits.gz')
sau.set_source('normgauss2d.source + const2d.background')
sau.set_stat('cstat')
# Ask for high-precision results
sau.set_method_opt('ftol', 1e-20)
sau.set_covar_opt('eps', 1e-20)
# Set start parameters close to simulation values to make the fit converge
sau.set_par('source.xpos', 101)
sau.set_par('source.ypos', 101)
sau.set_par('source.ampl', 1.1e3)
sau.set_par('source.fwhm', 10)
sau.set_par('background.c0', 1.1)
# Run fit and covariance estimation
# Results are automatically printed to the screen
sau.fit()
sau.covar()
# Sherpa uses fwhm instead of sigma as extension parameter ... need to convert
# http://cxc.harvard.edu/sherpa/ahelp/gauss2d.html
fwhm_to_sigma = 1. / np.sqrt(8 * np.log(2))
cov = sau.get_covar_results()
def test_beta_model():
tmpdir = tempfile.mkdtemp()
curdir = os.getcwd()
os.chdir(tmpdir)
r_c = 20.0
beta = 1.0
exp_time = Quantity(500.0, "ks")
e = spec.generate_energies(exp_time, area, prng=prng)
beta_src = BetaModel(ra0, dec0, r_c, beta, e.size, prng=prng)
write_photon_list("beta",
"beta",
e.flux,
beta_src.ra,
beta_src.dec,
e,
overwrite=True)
instrument_simulator("beta_simput.fits",
"beta_evt.fits",
exp_time,
"hdxi", [ra0, dec0],
ptsrc_bkgnd=False,
instr_bkgnd=False,
foreground=False,
prng=prng)
inst = get_instrument_from_registry("hdxi")
arf = AuxiliaryResponseFile(inst["arf"])
cspec = ConvolvedSpectrum(spec, arf)
ph_flux = cspec.get_flux_in_band(0.5, 7.0)[0].value
S0 = 3.0 * ph_flux / (2.0 * np.pi * r_c * r_c)
write_radial_profile("beta_evt.fits",
"beta_evt_profile.fits", [ra0, dec0],
0.0,
100.0,
200,
ctr_type="celestial",
emin=0.5,
emax=7.0,
overwrite=True)
load_data(1, "beta_evt_profile.fits", 3,
["RMID", "SUR_BRI", "SUR_BRI_ERR"])
set_stat("chi2")
set_method("levmar")
set_source("beta1d.src")
src.beta = 1.0
src.r0 = 10.0
src.ampl = 0.8 * S0
freeze(src.xpos)
fit()
set_covar_opt("sigma", 1.645)
covar()
res = get_covar_results()
assert np.abs(res.parvals[0] - r_c) < res.parmaxes[0]
assert np.abs(res.parvals[1] - beta) < res.parmaxes[1]
assert np.abs(res.parvals[2] - S0) < res.parmaxes[2]
os.chdir(curdir)
shutil.rmtree(tmpdir)
def test_chi2(self):
# Case 1: first ds has no error, second has, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2gehrels for ds1, chi2 for
# ds2, chi2gehrels for ds1,2
ui.load_data(1, self.data)
ui.load_data(2, self.data, use_errors=True)
ui.set_source(1, "gauss1d.g1")
ui.set_source(2, "gauss1d.g1")
ui.set_stat("chi2gehrels")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2gehrels', stat1)
self.assertEqual('chi2', stat2)
self.assertEqual('chi2gehrels', stat12)
# Case 2: first ds has errors, second has not, chi2-derived
# (chi2gehrels) statistic. I expect stat.name to be chi2 for ds1,
# chi2gehrels for ds2, chi2gehrels for ds1,2
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2gehrels', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2gehrels', stat12)
# Case 3: both datasets have errors, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2 for all of them.
ui.load_data(2, self.data, use_errors=True)
ui.load_data(1, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2', stat12)
# Case 4: first ds has errors, second has not, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('leastsq', stat2)
self.assertEqual('leastsq', stat1)
self.assertEqual('leastsq', stat12)
# Case 5: both ds have errors, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, self.data, use_errors=True)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('leastsq', stat2)
self.assertEqual('leastsq', stat1)
self.assertEqual('leastsq', stat12)
# Case 6: first ds has errors, second has not, CStat statistic
# I expect stat.name to be cstat for all of them.
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("cstat")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('cstat', stat2)
self.assertEqual('cstat', stat1)
self.assertEqual('cstat', stat12)
# Case7: select chi2 as statistic. One of the ds does not provide
# errors. I expect sherpa to raise a StatErr exception.
ui.set_stat('chi2')
caught = False
from sherpa.utils.err import StatErr
try:
ui.get_stat_info()
except StatErr:
caught = True
self.assertTrue(caught, msg='StatErr was not caught')
# Case8: select chi2 as statistic. Both datasets provide errors
# I expect stat to be 'chi2'
ui.load_data(2, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2', stat12)
def test_beta_model_flux():
tmpdir = tempfile.mkdtemp()
curdir = os.getcwd()
os.chdir(tmpdir)
r_c = 20.0
beta = 1.0
prng = 34
e = spec.generate_energies(exp_time, area, prng=prng)
beta_src = BetaModel(ra0, dec0, r_c, beta, e.size, prng=prng)
write_photon_list("beta",
"beta",
e.flux,
beta_src.ra,
beta_src.dec,
e,
overwrite=True)
instrument_simulator("beta_simput.fits",
"beta_evt.fits",
exp_time,
"acisi_cy0", [ra0, dec0],
ptsrc_bkgnd=False,
instr_bkgnd=False,
foreground=False,
roll_angle=37.0,
prng=prng)
ph_flux = spec.get_flux_in_band(0.5, 7.0)[0].value
S0 = 3.0 * ph_flux / (2.0 * np.pi * r_c * r_c)
wspec = spec.new_spec_from_band(0.5, 7.0)
make_exposure_map("beta_evt.fits",
"beta_expmap.fits",
wspec.emid.value,
weights=wspec.flux.value,
overwrite=True)
write_radial_profile("beta_evt.fits",
"beta_evt_profile.fits", [ra0, dec0],
0.0,
100.0,
200,
ctr_type="celestial",
emin=0.5,
emax=7.0,
expmap_file="beta_expmap.fits",
overwrite=True)
load_data(1, "beta_evt_profile.fits", 3,
["RMID", "SUR_FLUX", "SUR_FLUX_ERR"])
set_stat("chi2")
set_method("levmar")
set_source("beta1d.src")
src.beta = 1.0
src.r0 = 10.0
src.ampl = 0.8 * S0
freeze(src.xpos)
fit()
set_covar_opt("sigma", 1.645)
covar()
res = get_covar_results()
assert np.abs(res.parvals[0] - r_c) < res.parmaxes[0]
assert np.abs(res.parvals[1] - beta) < res.parmaxes[1]
assert np.abs(res.parvals[2] - S0) < res.parmaxes[2]
os.chdir(curdir)
shutil.rmtree(tmpdir)
def test_chi2(make_data_path, clean_astro_ui):
"bugs #11400, #13297, #12365"
data = make_data_path('3c273.pi')
# Case 1: first ds has no error, second has, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2gehrels for ds1, chi2 for
# ds2, chi2gehrels for ds1,2
ui.load_data(1, data)
ui.load_data(2, data, use_errors=True)
ui.set_source(1, "gauss1d.g1")
ui.set_source(2, "gauss1d.g1")
ui.set_stat("chi2gehrels")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'chi2gehrels'
assert stat2 == 'chi2'
assert stat12 == 'chi2gehrels'
# Case 2: first ds has errors, second has not, chi2-derived
# (chi2gehrels) statistic. I expect stat.name to be chi2 for ds1,
# chi2gehrels for ds2, chi2gehrels for ds1,2
ui.load_data(2, data)
ui.load_data(1, data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'chi2'
assert stat2 == 'chi2gehrels'
assert stat12 == 'chi2gehrels'
# Case 3: both datasets have errors, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2 for all of them.
ui.load_data(2, data, use_errors=True)
ui.load_data(1, data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'chi2'
assert stat2 == 'chi2'
assert stat12 == 'chi2'
# Case 4: first ds has errors, second has not, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, data)
ui.load_data(1, data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'leastsq'
assert stat2 == 'leastsq'
assert stat12 == 'leastsq'
# Case 5: both ds have errors, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, data, use_errors=True)
ui.load_data(1, data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'leastsq'
assert stat2 == 'leastsq'
assert stat12 == 'leastsq'
# Case 6: first ds has errors, second has not, CStat statistic
# I expect stat.name to be cstat for all of them.
ui.load_data(2, data)
ui.load_data(1, data, use_errors=True)
ui.set_stat("cstat")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'cstat'
assert stat2 == 'cstat'
assert stat12 == 'cstat'
# Case7: select chi2 as statistic. One of the ds does not provide
# errors. I expect sherpa to raise a StatErr exception.
ui.set_stat('chi2')
with pytest.raises(StatErr):
ui.get_stat_info()
# Case8: select chi2 as statistic. Both datasets provide errors
# I expect stat to be 'chi2'
ui.load_data(2, data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
assert stat1 == 'chi2'
assert stat2 == 'chi2'
assert stat12 == 'chi2'
def test_annulus():
tmpdir = tempfile.mkdtemp()
curdir = os.getcwd()
os.chdir(tmpdir)
r_in = 10.0
r_out = 30.0
e = spec.generate_energies(exp_time, area, prng=prng)
ann_src = AnnulusModel(ra0, dec0, r_in, r_out, e.size, prng=prng)
write_photon_list("ann",
"ann",
e.flux,
ann_src.ra,
ann_src.dec,
e,
overwrite=True)
instrument_simulator("ann_simput.fits",
"ann_evt.fits",
exp_time,
"hdxi", [ra0, dec0],
ptsrc_bkgnd=False,
instr_bkgnd=False,
foreground=False,
prng=prng)
inst = get_instrument_from_registry("hdxi")
arf = AuxiliaryResponseFile(inst["arf"])
cspec = ConvolvedSpectrum(spec, arf)
ph_flux = cspec.get_flux_in_band(0.5, 7.0)[0].value
S0 = ph_flux / (np.pi * (r_out**2 - r_in**2))
write_radial_profile("ann_evt.fits",
"ann_evt_profile.fits", [ra0, dec0],
1.1 * r_in,
0.9 * r_out,
100,
ctr_type="celestial",
emin=0.5,
emax=7.0,
overwrite=True)
load_data(1, "ann_evt_profile.fits", 3, ["RMID", "SUR_BRI", "SUR_BRI_ERR"])
set_stat("chi2")
set_method("levmar")
set_source("const1d.src")
src.c0 = 0.8 * S0
fit()
set_covar_opt("sigma", 1.645)
covar()
res = get_covar_results()
assert np.abs(res.parvals[0] - S0) < res.parmaxes[0]
os.chdir(curdir)
shutil.rmtree(tmpdir)
def test_chi2(self):
# Case 1: first ds has no error, second has, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2gehrels for ds1, chi2 for
# ds2, chi2gehrels for ds1,2
ui.load_data(1, self.data)
ui.load_data(2, self.data, use_errors=True)
ui.set_source(1, "gauss1d.g1")
ui.set_source(2, "gauss1d.g1")
ui.set_stat("chi2gehrels")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2gehrels', stat1)
self.assertEqual('chi2', stat2)
self.assertEqual('chi2gehrels', stat12)
# Case 2: first ds has errors, second has not, chi2-derived
# (chi2gehrels) statistic. I expect stat.name to be chi2 for ds1,
# chi2gehrels for ds2, chi2gehrels for ds1,2
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2gehrels', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2gehrels', stat12)
# Case 3: both datasets have errors, chi2-derived (chi2gehrels)
# statistic. I expect stat.name to be chi2 for all of them.
ui.load_data(2, self.data, use_errors=True)
ui.load_data(1, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2', stat12)
# Case 4: first ds has errors, second has not, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('leastsq', stat2)
self.assertEqual('leastsq', stat1)
self.assertEqual('leastsq', stat12)
# Case 5: both ds have errors, LeastSq statistic
# I expect stat.name to be leastsq for all of them.
ui.load_data(2, self.data, use_errors=True)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("leastsq")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('leastsq', stat2)
self.assertEqual('leastsq', stat1)
self.assertEqual('leastsq', stat12)
# Case 6: first ds has errors, second has not, CStat statistic
# I expect stat.name to be cstat for all of them.
ui.load_data(2, self.data)
ui.load_data(1, self.data, use_errors=True)
ui.set_stat("cstat")
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('cstat', stat2)
self.assertEqual('cstat', stat1)
self.assertEqual('cstat', stat12)
# Case7: select chi2 as statistic. One of the ds does not provide
# errors. I expect sherpa to raise a StatErr exception.
ui.set_stat('chi2')
caught = False
from sherpa.utils.err import StatErr
try:
ui.get_stat_info()
except StatErr:
caught = True
self.assertTrue(caught, msg='StatErr was not caught')
# Case8: select chi2 as statistic. Both datasets provide errors
# I expect stat to be 'chi2'
ui.load_data(2, self.data, use_errors=True)
si = ui.get_stat_info()
stat1 = si[0].statname
stat2 = si[1].statname
stat12 = si[2].statname
self.assertEqual('chi2', stat2)
self.assertEqual('chi2', stat1)
self.assertEqual('chi2', stat12)
# Licensed under a 3-clause BSD style license - see LICENSE.rst
"""Compute results with Sherpa"""
from __future__ import print_function, division
import numpy as np
import sherpa.astro.ui as sau
sau.load_data("counts.fits.gz")
sau.set_source("normgauss2d.source + const2d.background")
sau.set_stat("cstat")
# Ask for high-precision results
sau.set_method_opt("ftol", 1e-20)
sau.set_covar_opt("eps", 1e-20)
# Set start parameters close to simulation values to make the fit converge
sau.set_par("source.xpos", 101)
sau.set_par("source.ypos", 101)
sau.set_par("source.ampl", 1.1e3)
sau.set_par("source.fwhm", 10)
sau.set_par("background.c0", 1.1)
# Run fit and covariance estimation
# Results are automatically printed to the screen
sau.fit()
sau.covar()
# Sherpa uses fwhm instead of sigma as extension parameter ... need to convert
# http://cxc.harvard.edu/sherpa/ahelp/gauss2d.html
fwhm_to_sigma = 1.0 / np.sqrt(8 * np.log(2))
cov = sau.get_covar_results()
sigma = fwhm_to_sigma * cov.parvals[0]
sigma_err = fwhm_to_sigma * cov.parmaxes[0]
