def test_model_and_data(example_spec):
model, data = model_tools.model_and_data(example_spec)
assert model.spec["channels"] == example_spec["channels"]
assert model.config.modifier_settings == {
"normsys": {
"interpcode": "code4"
},
"histosys": {
"interpcode": "code4p"
},
}
assert data == [691, 1.0]
# requesting Asimov dataset
# TODO: request asimov dataset by setting asimove=True
# TODO: should return [112.429786, 1.0]
# TODO: Need to test overloaded method as well
# test handing a workspace instead of JSON
model, data = model_tools.model_and_data(pyhf.Workspace(example_spec))
assert model.spec["channels"] == example_spec["channels"]
assert model.config.modifier_settings == {
"normsys": {
"interpcode": "code4"
},
"histosys": {
"interpcode": "code4p"
},
}
assert data == [691, 1.0]
# without auxdata
model, data = model_tools.model_and_data(example_spec, with_aux=False)
assert data == [691]
def test_model_and_data(example_spec):
model, data = model_tools.model_and_data(example_spec)
assert model.spec["channels"] == example_spec["channels"]
assert model.config.modifier_settings == {
"normsys": {"interpcode": "code4"},
"histosys": {"interpcode": "code4p"},
}
assert data == [691, 1.0]
# requesting Asimov dataset
model, data = model_tools.model_and_data(pyhf.Workspace(example_spec), asimov=True)
assert model.spec["channels"] == example_spec["channels"]
assert model.config.modifier_settings == {
"normsys": {"interpcode": "code4"},
"histosys": {"interpcode": "code4p"},
}
assert data == [112.429786, 1.0]
# test handing a workspace instead of JSON
model, data = model_tools.model_and_data(pyhf.Workspace(example_spec))
assert model.spec["channels"] == example_spec["channels"]
assert model.config.modifier_settings == {
"normsys": {"interpcode": "code4"},
"histosys": {"interpcode": "code4p"},
}
assert data == [691, 1.0]
# without auxdata
model, data = model_tools.model_and_data(example_spec, include_auxdata=False)
assert data == [691]
def test_get_yields(example_spec):
model, data = model_tools.model_and_data(example_spec)
fit_results = fitter.fit(model, data)
y = yld.get_yields(example_spec, fit_results)
for region in y.regions:
assert np.allclose(y.yields[region], np.asarray([690.99844915]))
assert np.allclose(ak.to_list(y.uncertainties[region]),
[26.278787667809468])
assert np.allclose(y.data[region], np.asarray([691]))
assert y.regions == ['SR']
assert y.samples == ['signal']
def test_fit(mock_print, example_spec, example_spec_multibin):
model, data = model_tools.model_and_data(example_spec)
fit_results = fitter.fit(model, data)
mock_print.assert_called_once()
assert np.allclose(fit_results.bestfit, [1.1, 5.58731303])
assert np.allclose(fit_results.uncertainty, [0.0, 0.21248646])
assert fit_results.labels == ["staterror_SR", "mu_Sig"]
assert np.allclose(fit_results.best_twice_nll, 6.850287450660111)
assert np.allclose(fit_results.corr_mat, [[0.0, 0.0], [0.0, 1.0]])
# # TODO: Asimov fit, with fixed gamma (fixed not to Asimov MLE)
# model, data = model_tools.model_and_data(example_spec, asimov=True)
# fit_results = fitter.fit(model, data)
# # the gamma factor is multiplicative and fixed to 1.1, so the
# # signal strength needs to be 1/1.1 to compensate
# assert np.allclose(fit_results.bestfit, [1.1, 0.90917877])
# assert np.allclose(fit_results.uncertainty, [0.0, 0.12623179])
# assert fit_results.labels == ["staterror_SR", "mu_Sig"]
# assert np.allclose(fit_results.best_twice_nll, 5.68851093)
# assert np.allclose(fit_results.corr_mat, [[0.0, 0.0], [0.0, 1.0]])
# parameters held constant via keyword argument
model, data = model_tools.model_and_data(example_spec_multibin)
init_pars = model.config.suggested_init()
init_pars[0] = 0.9
init_pars[1] = 1.1
fixed_pars = model.config.suggested_fixed()
fixed_pars[0] = True
fixed_pars[1] = True
fit_results = fitter.fit(model,
data,
init_pars=init_pars,
fixed_pars=fixed_pars)
assert np.allclose(fit_results.bestfit, [0.9, 1.1, 1.11996446, 0.96618774])
assert np.allclose(fit_results.uncertainty,
[0.0, 0.0, 0.1476617, 0.17227148])
assert np.allclose(fit_results.best_twice_nll, 11.2732492)