class ComprehensiveTest(object):
""" Perform the likelihood ratio test using a "comprehensive model"
Usage:
roi = ...
test = ComprehensiveTest(roi, 'source_name', PowerLaw(), DMFitFunction(),
fit_kwargs=dict(use_gradient=False))
print test.todict()
test.plot('test.pdf')
"""
defaults = (
('fit_kwargs', dict(), 'Parameters to pass into the roi.fit function'),
('localize', False, 'Localize source with both hypothesis'),
('quiet', False, 'Shut this code up'),
)
@keyword_options.decorate(defaults)
def __init__(self, roi, which, model0, model1, **kwargs):
keyword_options.process(self, kwargs)
state = PointlikeState(roi)
self.roi = roi
self.which = which
self.model0 = model0.copy()
self.model1 = model1.copy()
self.compute()
state.restore()
def fit(self):
roi = self.roi
which = self.which
fit_prefactor(roi, which, **self.fit_kwargs)
roi.fit(**self.fit_kwargs)
if self.localize:
roi.localize(which=which, update=True)
roi.fit(**self.fit_kwargs)
ll = -roi.logLikelihood(roi.parameters())
return ll
def compute(self):
roi = self.roi
which = self.which
roi.modify(which=which, model=self.model0, keep_old_flux=False)
if not self.quiet:
print 'fit source %s with model0 (%s)' % (which, self.model0.name)
self.ll_0 = self.fit()
if not self.quiet:
print roi
sed = PointlikeSED(roi, name=which)
self.sed_points = sed.todict()
# TODO, get SED points
emin, emax = get_full_energy_range(roi)
flux = self.model0.i_flux(emin=emin, emax=emax)
self.model1.set_flux(flux, emin=emin, emax=emax)
self.comprehensive_model = ComprehensiveModel(self.model0.copy(),
self.model1.copy())
self.comprehensive_model.theta = 0.5
self.comprehensive_model_start = self.comprehensive_model.copy()
if not self.quiet:
print 'fit source %s with model1 (%s)' % (which, self.model1.name)
roi.modify(which=self.which,
model=self.comprehensive_model,
keep_old_flux=False)
self.ll_1 = self.fit()
if not self.quiet:
print roi
self.TS_comp = 2 * (self.ll_1 - self.ll_0)
def todict(self):
return tolist(
dict(
sed_points=self.sed_points,
model0=spectrum_to_dict(self.model0),
model1=spectrum_to_dict(self.comprehensive_model),
TS_comp=self.TS_comp,
ll_1=self.ll_1,
ll_0=self.ll_0,
))
def plot(self, filename):
sed = SED(self.sed_points)
# Plot SED points
axes = sed.plot()
# Plot model0, model1
sed.plot_spectrum(self.model0, axes=axes, label='model0')
sed.plot_spectrum(self.comprehensive_model, axes=axes, label='model1')
sed.plot_spectrum(self.comprehensive_model_start,
axes=axes,
label='model1 (start)')
axes.legend()
P.savefig(filename)
class ComprehensiveTest(object):
""" Perform the likelihood ratio test using a "comprehensive model"
Usage:
roi = ...
test = ComprehensiveTest(roi, 'source_name', PowerLaw(), DMFitFunction(),
fit_kwargs=dict(use_gradient=False))
print test.todict()
test.plot('test.pdf')
"""
defaults = (
('fit_kwargs', dict(), 'Parameters to pass into the roi.fit function'),
('localize', False, 'Localize source with both hypothesis'),
('quiet', False, 'Shut this code up'),
)
@keyword_options.decorate(defaults)
def __init__(self, roi, which, model0, model1, **kwargs):
keyword_options.process(self, kwargs)
state = PointlikeState(roi)
self.roi=roi
self.which=which
self.model0=model0.copy()
self.model1=model1.copy()
self.compute()
state.restore()
def fit(self):
roi=self.roi
which=self.which
fit_prefactor(roi, which, **self.fit_kwargs)
roi.fit(**self.fit_kwargs)
if self.localize:
roi.localize(which=which, update=True)
roi.fit(**self.fit_kwargs)
ll=-roi.logLikelihood(roi.parameters())
return ll
def compute(self):
roi=self.roi
which=self.which
roi.modify(which=which, model=self.model0, keep_old_flux=False)
if not self.quiet: print 'fit source %s with model0 (%s)' % (which,self.model0.name)
self.ll_0 = self.fit()
if not self.quiet:
print roi
sed = PointlikeSED(roi,name=which)
self.sed_points = sed.todict()
# TODO, get SED points
emin,emax=get_full_energy_range(roi)
flux=self.model0.i_flux(emin=emin,emax=emax)
self.model1.set_flux(flux,emin=emin,emax=emax)
self.comprehensive_model = ComprehensiveModel(self.model0.copy(), self.model1.copy())
self.comprehensive_model.theta=0.5
self.comprehensive_model_start = self.comprehensive_model.copy()
if not self.quiet: print 'fit source %s with model1 (%s)' % (which,self.model1.name)
roi.modify(which=self.which, model=self.comprehensive_model, keep_old_flux=False)
self.ll_1 = self.fit()
if not self.quiet:
print roi
self.TS_comp = 2*(self.ll_1-self.ll_0)
def todict(self):
return tolist(dict(
sed_points = self.sed_points,
model0 = spectrum_to_dict(self.model0),
model1 = spectrum_to_dict(self.comprehensive_model),
TS_comp = self.TS_comp,
ll_1 = self.ll_1,
ll_0 = self.ll_0,
))
def plot(self, filename):
sed = SED(self.sed_points)
# Plot SED points
axes=sed.plot()
# Plot model0, model1
sed.plot_spectrum(self.model0, axes=axes, label='model0')
sed.plot_spectrum(self.comprehensive_model, axes=axes, label='model1')
sed.plot_spectrum(self.comprehensive_model_start, axes=axes, label='model1 (start)')
axes.legend()
P.savefig(filename)