def test_compute_criteria_computes_criteria_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
test_aic = 169440.83719024697
test_bic = 169245.62163088709
test_deviance = 337950.6823459795
assert np.isclose(res.aic, test_aic, atol=0.1, rtol=0.1)
assert np.isclose(res.bic, test_bic, atol=0.1, rtol=0.1)
assert np.isclose(res.deviance, test_deviance, atol=0.1, rtol=0.1)
def test_object_has_right_attributes(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
assert hasattr(res, "p_opt")
assert hasattr(res, "result")
assert hasattr(res, "deviance")
assert hasattr(res, "aic")
assert hasattr(res, "bic")
assert hasattr(res, "model")
assert isinstance(res.model, models.Const1D)
def test_compute_criteria_works_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
test_aic = res.result + 2.0 * res.p_opt.shape[0]
test_bic = res.result + res.p_opt.shape[0] * \
np.log(self.lpost.x.shape[0])
test_deviance = -2 * self.lpost.loglikelihood(res.p_opt, neg=False)
assert np.isclose(res.aic, test_aic, atol=0.1, rtol=0.1)
assert np.isclose(res.bic, test_bic, atol=0.1, rtol=0.1)
assert np.isclose(res.deviance, test_deviance, atol=0.1, rtol=0.1)
def test_object_initializes_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
assert hasattr(res, "p_opt")
assert hasattr(res, "result")
assert hasattr(res, "deviance")
assert hasattr(res, "aic")
assert hasattr(res, "bic")
assert hasattr(res, "model")
assert isinstance(res.model, models.Const1D)
assert res.p_opt == self.opt.x, "res.p_opt must be the same as opt.x!"
assert np.isclose(res.p_opt[0], 2.0, atol=0.1, rtol=0.1)
assert res.model == self.lpost.model
assert res.result == self.opt.fun
mean_model = np.ones_like(self.lpost.x) * self.opt.x[0]
assert np.allclose(res.mfit, mean_model), "res.model should be exactly " \
"the model for the data."
def test_merit_calculated_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
test_merit = np.sum(((self.ps.power - 2.0)/2.0)**2.)
assert np.isclose(res.merit, test_merit, rtol=0.2)
def test_merit_calculated_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
test_merit = 98770.654981073574
assert np.isclose(res.merit, test_merit, atol=0.1, rtol=0.1)
def test_compute_model_works_correctly(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
mean_model = np.ones_like(self.lpost.x) * self.opt.x[0]
assert np.all(
res.mfit == mean_model), "res.model should be exactly " \
"the model for the data."
def test_result_is_same_as_in_opt(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
assert res.result == self.opt.fun
def test_model_is_same_as_in_lpost(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
assert res.model == self.lpost.model
def test_p_opt_is_correct(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)
assert res.p_opt == self.opt.x, "res.p_opt must be the same as opt.x!"
assert np.isclose(res.p_opt[0], 2.0, atol=0.1, rtol=0.1)
def test_object_initializes(self):
res = OptimizationResults(self.lpost, self.opt, neg=self.neg)