def test_first_hill(self):
# typical bimodal distribution
data = np.concatenate([self.dist_norm1, self.dist_norm2])[:,
np.newaxis]
estimator = KernelDensity(kernel='gaussian', bandwidth=0.5)
kde = estimator.fit(data)
x, y = Analyze.density_func(data, kde, 100)
obs_x, obs_y = Analyze.first_hill(x, y)
exp_x, exp_y = 1.0971012583068704, 2.5302323352207674
self.assertAlmostEqual(obs_x, exp_x)
self.assertAlmostEqual(obs_y, exp_y)
# peak larger than valley
data = np.negative(data)
kde = estimator.fit(data)
x, y = Analyze.density_func(data, kde, 100)
with self.assertRaises(ValueError) as ctx:
Analyze.first_hill(x, y)
msg = 'Peak is larger than valley.'
self.assertEqual(str(ctx.exception), msg)
# unimodal distribution
data = self.dist_norm1[:, np.newaxis]
kde = estimator.fit(data)
x, y = Analyze.density_func(data, kde, 100)
with self.assertRaises(ValueError) as ctx:
Analyze.first_hill(x, y)
msg = 'Cannot identify at least two peaks.'
self.assertEqual(str(ctx.exception), msg)
def test_plot_density(self):
data = np.concatenate([self.dist_norm1, self.dist_norm2])[:,
np.newaxis]
estimator = KernelDensity(kernel='gaussian', bandwidth=0.5)
kde = estimator.fit(data)
x, y = Analyze.density_func(data, kde, 100)
peak, valley = Analyze.first_hill(x, y)
th = valley - (valley - peak) * 0.5 / 100
fp = join(self.tmpdir, 'tmp.png')
Analyze.plot_density(x, y, peak, valley, th, fp)
self.assertTrue(isfile(fp))
remove(fp)
def test_density_func(self):
data = self.dist_norm1[:, np.newaxis]
estimator = KernelDensity(kernel='gaussian', bandwidth=0.5)
kde = estimator.fit(data)
obs = Analyze.density_func(data, kde, 10)
exp = (np.array([
-1.48253468, 0.0939095, 1.67035369, 3.24679787, 4.82324206,
6.39968624, 7.97613043, 9.55257461, 11.1290188, 12.70546298
]),
np.array([
0.00104342, 0.00788705, 0.0496806, 0.13173376, 0.19176352,
0.15754466, 0.06992292, 0.02140856, 0.00150463, 0.00053637
]))
np.testing.assert_array_almost_equal(obs, exp)
