def test_euclidean_perfect(self):
observed = compare_sinks(self.mpm1, self.mpm1, 'euclidean')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Euclidean distance'])
assert_data_frame_almost_equal(observed, expected)
def test_euclidean(self):
mpm1 = {
'Unknown': {
'sink1': 0.25
},
'Source1': {
'sink1': 0.50
},
'Source2': {
'sink1': 0.25
}
}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {
'Unknown': {
'sink1': 0.1
},
'Source1': {
'sink1': 0.1
},
'Source2': {
'sink1': 0.8
}
}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'euclidean')
expected_ids = ['sink1']
# expected values computed by calling
# scipy.stats.spatial.distance.euclidean directly
expected_values = [0.6964194]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Euclidean distance'])
assert_data_frame_almost_equal(observed, expected)
def test_absolute_difference(self):
mpm1 = {
'Unknown': {
'sink1': 0.25
},
'Source1': {
'sink1': 0.50
},
'Source2': {
'sink1': 0.25
}
}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {
'Unknown': {
'sink1': 0.1
},
'Source2': {
'sink1': 0.8
},
'Source1': {
'sink1': 0.1
}
}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'absolute_difference')
expected_ids = ['sink1']
# expected values computed by hand
expected_values = [(0.4, 0.55, 0.15)]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Source1', 'Source2', 'Unknown'])
assert_data_frame_almost_equal(observed.sort_index(axis=1),
expected.sort_index(axis=1))
def test_pearsonr(self):
mpm1 = {
'Unknown': {
'sink1': 0.25
},
'Source1': {
'sink1': 0.50
},
'Source2': {
'sink1': 0.25
}
}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {
'Unknown': {
'sink1': 0.1
},
'Source1': {
'sink1': 0.1
},
'Source2': {
'sink1': 0.8
}
}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'pearson')
expected_ids = ['sink1']
# expected values computed by calling scipy.stats.pearsonr directly
expected_values = [(-0.5, 2. / 3)]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Pearson r', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_pearson_perfect(self):
observed = compare_sinks(self.mpm1, self.mpm1, 'pearson')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [(1.0, 0.0), (1.0, 0.0), (1.0, 0.0), (1.0, 0.0),
(1.0, 0.0), (1.0, 0.0)]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Pearson r', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_euclidean_perfect(self):
observed = compare_sinks(self.mpm1, self.mpm1, 'euclidean')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Euclidean distance'])
assert_data_frame_almost_equal(observed, expected)
def test_pearson_perfect(self):
observed = compare_sinks(self.mpm1, self.mpm1, 'pearson')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [(1.0, 0.0), (1.0, 0.0), (1.0, 0.0), (1.0, 0.0),
(1.0, 0.0), (1.0, 0.0)]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Pearson r', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_order_independence_sources(self):
mpm1 = self.mpm1.sort_index(ascending=False, axis=1)
mpm2 = self.mpm1.copy().sort_index(ascending=True, axis=1)
# confirm that the columns are now different
self.assertEqual(list(mpm1.columns), list(reversed(mpm2.columns)))
observed = compare_sinks(self.mpm1, self.mpm1, 'spearman')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [(1.0, 0.0), (1.0, 0.0), (1.0, 0.0), (1.0, 0.0),
(1.0, 0.0), (1.0, 0.0)]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Spearman rho', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_order_independence_sources(self):
mpm1 = self.mpm1.sort_index(ascending=False, axis=1)
mpm2 = self.mpm1.copy().sort_index(ascending=True, axis=1)
# confirm that the columns are now different
self.assertEqual(list(mpm1.columns), list(reversed(mpm2.columns)))
observed = compare_sinks(self.mpm1, self.mpm1, 'spearman')
expected_ids = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink6']
expected_values = [(1.0, 0.0), (1.0, 0.0), (1.0, 0.0), (1.0, 0.0),
(1.0, 0.0), (1.0, 0.0)]
expected = pd.DataFrame(expected_values,
index=expected_ids,
columns=['Spearman rho', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_pearsonr(self):
mpm1 = {'Unknown': {'sink1': 0.25},
'Source1': {'sink1': 0.50},
'Source2': {'sink1': 0.25}}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {'Unknown': {'sink1': 0.1},
'Source1': {'sink1': 0.1},
'Source2': {'sink1': 0.8}}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'pearson')
expected_ids = ['sink1']
# expected values computed by calling scipy.stats.pearsonr directly
expected_values = [(-0.5, 2./3)]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Pearson r', 'p'])
assert_data_frame_almost_equal(observed, expected)
def test_absolute_difference(self):
mpm1 = {'Unknown': {'sink1': 0.25},
'Source1': {'sink1': 0.50},
'Source2': {'sink1': 0.25}}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {'Unknown': {'sink1': 0.1},
'Source2': {'sink1': 0.8},
'Source1': {'sink1': 0.1}}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'absolute_difference')
expected_ids = ['sink1']
# expected values computed by hand
expected_values = [(0.4, 0.55, 0.15)]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Source1', 'Source2', 'Unknown'])
assert_data_frame_almost_equal(observed, expected)
def test_euclidean(self):
mpm1 = {'Unknown': {'sink1': 0.25},
'Source1': {'sink1': 0.50},
'Source2': {'sink1': 0.25}}
mpm1 = pd.DataFrame(mpm1)
mpm2 = {'Unknown': {'sink1': 0.1},
'Source1': {'sink1': 0.1},
'Source2': {'sink1': 0.8}}
mpm2 = pd.DataFrame(mpm2)
observed = compare_sinks(mpm1, mpm2, 'euclidean')
expected_ids = ['sink1']
# expected values computed by calling
# scipy.stats.spatial.distance.euclidean directly
expected_values = [0.6964194]
expected = pd.DataFrame(expected_values, index=expected_ids,
columns=['Euclidean distance'])
assert_data_frame_almost_equal(observed, expected)
def test_non_overlapping_sources(self):
mpm2 = self.mpm1.copy()
mpm2.columns = ['source1', 'source2', 'source4', 'Unknown']
with self.assertRaisesRegex(ValueError, 'Sources'):
compare_sinks(self.mpm1, mpm2, 'spearman')
def test_non_overlapping_sinks(self):
mpm2 = self.mpm1.copy()
mpm2.index = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink7']
with self.assertRaisesRegex(ValueError, 'Sinks'):
compare_sinks(self.mpm1, mpm2, 'spearman')
def test_non_overlapping_sinks(self):
mpm2 = self.mpm1.copy()
mpm2.index = ['sink1', 'sink2', 'sink3', 'sink4', 'sink5', 'sink7']
with self.assertRaisesRegex(ValueError, 'Sinks'):
compare_sinks(self.mpm1, mpm2, 'spearman')
def test_non_overlapping_sources(self):
mpm2 = self.mpm1.copy()
mpm2.columns = ['source1', 'source2', 'source4', 'Unknown']
with self.assertRaisesRegex(ValueError, 'Sources'):
compare_sinks(self.mpm1, mpm2, 'spearman')
def test_unknown_metric(self):
with self.assertRaises(KeyError):
compare_sinks(self.mpm1, self.mpm1, 'not-a-metric')
def test_unknown_metric(self):
with self.assertRaises(KeyError):
compare_sinks(self.mpm1, self.mpm1, 'not-a-metric')