def test_numba_likelihood(self):
"""
Test likelihood using numba against the standard likelihood.
"""
het = HeterodynedData(self.data,
times=self.times,
detector=self.detector,
par=self.parfile)
priors = dict()
priors["h0"] = Uniform(0.0, 1.0e-23, "h0")
for likelihood in ["gaussian", "studentst"]:
like1 = TargetedPulsarLikelihood(het,
PriorDict(priors),
likelihood=likelihood)
like1.parameters = {"h0": 1e-24}
like2 = TargetedPulsarLikelihood(het,
PriorDict(priors),
likelihood=likelihood,
numba=True)
like2.parameters = {"h0": 1e-24}
assert like1.log_likelihood() == like2.log_likelihood()
def test_includephase_likelihood(self):
"""
Test the likelihood when include phase is set to True.
"""
het = HeterodynedData(self.data,
times=self.times,
detector=self.detector,
par=self.parfile)
priors = dict()
priors["h0"] = Uniform(0.0, 1.0e-23, "h0")
# run with includephase as False
like1 = TargetedPulsarLikelihood(het,
PriorDict(priors),
likelihood="studentst")
like1.parameters = {"h0": 1e-24}
logl1 = like1.log_likelihood()
# set includephase to True
like2 = TargetedPulsarLikelihood(het,
PriorDict(priors),
likelihood="studentst")
like2.parameters = {"h0": 1e-24}
like2.include_phase = True
logl2 = like2.log_likelihood()
print(f"{logl1:.15f} {logl2:.15f}")
assert np.allclose([logl1], [logl2], atol=1e-10, rtol=0.0)
def test_likelihood_null_likelihood(self):
"""
Test likelihood and null likelihood.
"""
het = HeterodynedData(self.data,
times=self.times,
detector=self.detector,
par=self.parfile)
priors = dict()
priors["h0"] = Uniform(0.0, 1.0e-23, "h0")
for likelihood in ["gaussian", "studentst"]:
like = TargetedPulsarLikelihood(het,
PriorDict(priors),
likelihood=likelihood)
like.parameters = {"h0": 0.0}
assert like.log_likelihood() == like.noise_log_likelihood()
# create some fake heterodyned data
detector = "H1" # the detector to use
times = np.linspace(1000000000.0, 1000086340.0, 1440) # times
het = HeterodynedData(
times=times,
inject=True,
par=parfile,
injpar=parfile,
fakeasd=1e-24,
detector=detector,
)
# set prior on h0
priors = dict()
priors["h0"] = Uniform(0.0, 1e-24, "h0")
like = TargetedPulsarLikelihood(het, PriorDict(priors))
N = 500
h0s = np.linspace(0.0, 1e-24, N)
post = np.zeros(N)
for i, h0 in enumerate(h0s):
params = {"h0": h0}
like.parameters = params
post[i] = like.log_likelihood() + priors["h0"].ln_prob(h0)
pl.plot(h0s, np.exp(post - np.max(post)), "b")
pl.axvline(het.injpar["H0"])
pl.show()