def test_profiler(self):
# defining some dummy parameters and observables
Parameter('tmp a')
Parameter('tmp b')
Parameter('tmp c')
Parameter('tmp d')
p = ParameterConstraints()
p.set_constraint('tmp b', '2+-0.3')
p.set_constraint('tmp c', '0.2+-0.1')
p.set_constraint('tmp d', '1+-0.5')
def prediction(wc_obj, par):
return par['tmp a']**2 + par['tmp b'] + par['tmp c'] + par[
'tmp d']**2
flavio.Observable('tmp obs')
Prediction('tmp obs', prediction)
m = Measurement('tmp measurement')
m.add_constraint(['tmp obs'],
flavio.statistics.probability.NormalDistribution(
1, 0.2))
# test 1D profiler
fit_1d = FrequentistFit('test profiler 1d', p, ['tmp a'],
['tmp b', 'tmp c', 'tmp d'], ['tmp obs'])
profiler_1d = profiler.Profiler1D(fit_1d, -10, 10)
x, z, n = profiler_1d.run(steps=4)
self.assertEqual(x.shape, (4, ))
self.assertEqual(z.shape, (4, ))
self.assertEqual(n.shape, (3, 4))
npt.assert_array_equal(x, profiler_1d.x)
npt.assert_array_equal(z, profiler_1d.log_profile_likelihood)
npt.assert_array_equal(n, profiler_1d.profile_nuisance)
pdat = profiler_1d.pvalue_prob_plotdata()
npt.assert_array_equal(pdat['x'], x)
# test 2D profiler
p.remove_constraint('d')
fit_2d = FrequentistFit('test profiler 2d', p, ['tmp a', 'tmp d'],
['tmp b', 'tmp c'], ['tmp obs'])
profiler_2d = profiler.Profiler2D(fit_2d, -10, 10, -10, 10)
x, y, z, n = profiler_2d.run(steps=(3, 4))
self.assertEqual(x.shape, (3, ))
self.assertEqual(y.shape, (4, ))
self.assertEqual(z.shape, (3, 4))
self.assertEqual(n.shape, (2, 3, 4))
npt.assert_array_equal(x, profiler_2d.x)
npt.assert_array_equal(y, profiler_2d.y)
npt.assert_array_equal(z, profiler_2d.log_profile_likelihood)
npt.assert_array_equal(n, profiler_2d.profile_nuisance)
pdat = profiler_2d.contour_plotdata()
npt.assert_array_almost_equal(pdat['z'], -2 * (z - np.max(z)))
# delete dummy instances
for p in ['tmp a', 'tmp b', 'tmp c', 'tmp d']:
Parameter.del_instance(p)
FrequentistFit.del_instance('test profiler 1d')
Observable.del_instance('tmp obs')
Measurement.del_instance('tmp measurement')
def test_sm_covariance(self):
o1 = Observable('test_obs 1')
o2 = Observable('test_obs 2')
def f1(wc_obj, par_dict):
return par_dict['m_b']
def f2(wc_obj, par_dict):
return par_dict['m_c']
Prediction('test_obs 1', f1)
Prediction('test_obs 2', f2)
cov_par = np.array([[0.1**2, 0.1 * 0.2 * 0.3],
[0.1 * 0.2 * 0.3, 0.2**2]])
d = flavio.statistics.probability.MultivariateNormalDistribution(
[4.2, 1.2], covariance=cov_par)
par = copy.deepcopy(flavio.parameters.default_parameters)
par.add_constraint(['m_b', 'm_c'], d)
# test serial
np.random.seed(135)
cov = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000,
par_vary='all',
par_obj=par)
npt.assert_array_almost_equal(cov, cov_par, decimal=2)
# test parallel
np.random.seed(135)
cov_parallel = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000,
par_vary='all',
par_obj=par,
threads=4)
npt.assert_array_almost_equal(cov, cov_parallel, decimal=6)
np.random.seed(135)
cov_1 = flavio.sm_covariance(['test_obs 1'],
N=1000,
par_vary='all',
par_obj=par)
# test with single observable
npt.assert_array_almost_equal(cov_1, cov[0, 0])
# test with fixed parameter
cov_f = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000,
par_vary=['m_b'],
par_obj=par)
npt.assert_array_almost_equal(cov_f, [[cov_par[0, 0], 0], [0, 0]],
decimal=3)
# delete dummy instances
Observable.del_instance('test_obs 1')
Observable.del_instance('test_obs 2')
def test_functions(self):
o = Observable('test_obs')
o.arguments = ['x']
def f(wc_obj, par_dict, x):
return x
pr = Prediction('test_obs', f )
wc_obj = None
self.assertEqual(flavio.sm_prediction('test_obs', 7), 7)
self.assertEqual(flavio.np_prediction('test_obs', x=7, wc_obj=wc_obj), 7)
self.assertEqual(flavio.sm_uncertainty('test_obs', 7), 0)
self.assertEqual(flavio.np_uncertainty('test_obs', x=7, wc_obj=wc_obj), 0)
self.assertEqual(flavio.sm_uncertainty('test_obs', 7, threads=2), 0)
self.assertEqual(flavio.np_uncertainty('test_obs', x=7, wc_obj=wc_obj, threads=2), 0)
# delete dummy instance
Observable.del_instance('test_obs')
def test_exp_combo(self):
o = Observable('test_obs')
o.arguments = ['x']
m = Measurement('test_obs measurement 1')
m.add_constraint([('test_obs', 1)], MultivariateNormalDistribution([1, 2], np.eye(2)))
# error: no measurement
with self.assertRaises(ValueError):
flavio.combine_measurements('test_obs', x=1, include_measurements=['bla'])
m.add_constraint([('test_obs', 1)], NormalDistribution(2, 3))
combo = flavio.combine_measurements('test_obs', x=1)
self.assertEqual(combo.central_value, 2)
self.assertEqual(combo.standard_deviation, 3)
m2 = Measurement('test_obs measurement 2')
m2.add_constraint([('test_obs', 1)], NormalDistribution(3, 3))
combo = flavio.combine_measurements('test_obs', x=1)
self.assertAlmostEqual(combo.central_value, 2.5)
self.assertAlmostEqual(combo.standard_deviation, sqrt(9 / 2))
Observable.del_instance('test_obs')
def test_sm_covariance(self):
o1 = Observable( 'test_obs 1' )
o2 = Observable( 'test_obs 2' )
def f1(wc_obj, par_dict):
return par_dict['m_b']
def f2(wc_obj, par_dict):
return par_dict['m_c']
Prediction('test_obs 1', f1)
Prediction('test_obs 2', f2)
cov_par = np.array([[0.1**2, 0.1*0.2*0.3], [0.1*0.2*0.3, 0.2**2]])
d = flavio.statistics.probability.MultivariateNormalDistribution([4.2, 1.2], covariance=cov_par)
par = copy.deepcopy(flavio.parameters.default_parameters)
par.add_constraint(['m_b', 'm_c'], d)
# test serial
np.random.seed(135)
cov = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000, par_vary='all', par_obj=par)
npt.assert_array_almost_equal(cov, cov_par, decimal=2)
# test parallel
np.random.seed(135)
cov_parallel = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000, par_vary='all', par_obj=par,
threads=4)
npt.assert_array_almost_equal(cov, cov_parallel, decimal=6)
np.random.seed(135)
cov_1 = flavio.sm_covariance(['test_obs 1'],
N=1000, par_vary='all', par_obj=par)
# test with single observable
npt.assert_array_almost_equal(cov_1, cov[0, 0])
# test with fixed parameter
cov_f = flavio.sm_covariance(['test_obs 1', 'test_obs 2'],
N=1000, par_vary=['m_b'], par_obj=par)
npt.assert_array_almost_equal(cov_f, [[cov_par[0, 0], 0], [0, 0]], decimal=3)
# delete dummy instances
Observable.del_instance('test_obs 1')
Observable.del_instance('test_obs 2')
def test_profiler(self):
# defining some dummy parameters and observables
Parameter('tmp a');
Parameter('tmp b');
Parameter('tmp c');
Parameter('tmp d');
p = ParameterConstraints()
p.set_constraint('tmp b', '2+-0.3')
p.set_constraint('tmp c', '0.2+-0.1')
p.set_constraint('tmp d', '1+-0.5')
def prediction(wc_obj, par):
return par['tmp a']**2+par['tmp b']+par['tmp c']+par['tmp d']**2
flavio.Observable('tmp obs');
Prediction('tmp obs', prediction);
m=Measurement('tmp measurement')
m.add_constraint(['tmp obs'],
flavio.statistics.probability.NormalDistribution(1, 0.2))
# test 1D profiler
fit_1d = FrequentistFit('test profiler 1d',
p, ['tmp a'], ['tmp b', 'tmp c', 'tmp d'], ['tmp obs'])
profiler_1d = profiler.Profiler1D(fit_1d, -10, 10)
x, z, n = profiler_1d.run(steps=4)
self.assertEqual(x.shape, (4,))
self.assertEqual(z.shape, (4,))
self.assertEqual(n.shape, (3, 4))
npt.assert_array_equal(x, profiler_1d.x)
npt.assert_array_equal(z, profiler_1d.log_profile_likelihood)
npt.assert_array_equal(n, profiler_1d.profile_nuisance)
pdat = profiler_1d.pvalue_prob_plotdata()
npt.assert_array_equal(pdat['x'], x)
# test multiprocessing
for threads in [2, 3, 4]:
xt, zt, nt = profiler_1d.run(steps=4, threads=threads)
npt.assert_array_almost_equal(x, xt, decimal=4)
npt.assert_array_almost_equal(z, zt, decimal=4)
npt.assert_array_almost_equal(n, nt, decimal=4)
with self.assertRaises(ValueError):
profiler_1d.run(steps=4, threads=5)
# test 2D profiler
p.remove_constraint('d')
fit_2d = FrequentistFit('test profiler 2d',
p, ['tmp a', 'tmp d'], ['tmp b', 'tmp c'], ['tmp obs'])
profiler_2d = profiler.Profiler2D(fit_2d, -10, 10, -10, 10)
x, y, z, n = profiler_2d.run(steps=(3,4))
self.assertEqual(x.shape, (3,))
self.assertEqual(y.shape, (4,))
self.assertEqual(z.shape, (3, 4))
self.assertEqual(n.shape, (2, 3, 4))
npt.assert_array_equal(x, profiler_2d.x)
npt.assert_array_equal(y, profiler_2d.y)
npt.assert_array_equal(z, profiler_2d.log_profile_likelihood)
npt.assert_array_equal(n, profiler_2d.profile_nuisance)
pdat = profiler_2d.contour_plotdata()
npt.assert_array_almost_equal(pdat['z'], -2*(z-np.max(z)))
# test multiprocessing
for threads in [2, 5, 12]:
xt, yt, zt, nt = profiler_2d.run(steps=(3,4))
npt.assert_array_almost_equal(x, xt, decimal=4)
npt.assert_array_almost_equal(y, yt, decimal=4)
npt.assert_array_almost_equal(z, zt, decimal=4)
npt.assert_array_almost_equal(n, nt, decimal=4)
with self.assertRaises(ValueError):
profiler_2d.run(steps=(3,4), threads=13)
# delete dummy instances
for p in ['tmp a', 'tmp b', 'tmp c', 'tmp d']:
Parameter.del_instance(p)
FrequentistFit.del_instance('test profiler 1d')
Observable.del_instance('tmp obs')
Measurement.del_instance('tmp measurement')
