def test_bhattacharyya_no_column_not_in_both_crossover_cat_to_cont(self):
cat = ['cat_same_dist', 'cat_diff_dist', 'cat_not_in_both']
cont = ['cont_same_dist', 'cont_diff_dist']
d = sf.bhattacharyya(self.df_1, self.df_2, categorical_variables=cat, continuous_variables=cont)
self.assertTrue(d.isnull()['cat_not_in_both'])
self.assertTrue((~d.isnull().drop('cat_not_in_both')).all())
def test_bhattacharyya_cat_distance(self):
var = ['cat_same_dist', 'cat_diff_dist']
d = sf.bhattacharyya(self.df_1, self.df_2, categorical_variables=var)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(all([v in d.index for v in var]+[v in var for v in d.index]))
self.assertEqual(0.0, d['cat_same_dist'])
self.assertEqual(0.0614, np.round(d['cat_diff_dist'], decimals=4))
def test_bhattacharyya_cont_distance(self):
var = ['cont_same_dist', 'cont_diff_dist']
d = sf.bhattacharyya(self.df_1, self.df_2, continuous_variables=var)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(all([v in d.index for v in var]+[v in var for v in d.index]))
self.assertEqual(0.0138, np.round(d['cont_same_dist'], decimals=4))
self.assertEqual(0.4205, np.round(d['cont_diff_dist'], decimals=4))
def test_bhattacharyya_full_calculation(self):
cat = ['cat_same_dist', 'cat_diff_dist']
cont = ['cont_same_dist', 'cont_diff_dist']
d = sf.bhattacharyya(self.df_1, self.df_2, categorical_variables=cat, continuous_variables=cont)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(all([v in d.index for v in cont]
+ [(v in cat) or (v in cont) for v in d.index]
+ [v in d.index for v in cat]))
self.assertEqual(0.0, d['cat_same_dist'])
self.assertEqual(0.0614, np.round(d['cat_diff_dist'], decimals=4))
self.assertEqual(0.0138, np.round(d['cont_same_dist'], decimals=4))
self.assertEqual(0.4205, np.round(d['cont_diff_dist'], decimals=4))
def test_bhattacharyya_cat_column_not_in_both(self):
var = ['cat_same_dist', 'cat_diff_dist', 'cat_not_in_both']
d = sf.bhattacharyya(self.df_1, self.df_2, categorical_variables=var)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(all([v in d.index for v in var] + [v in var for v in d.index]))
self.assertTrue(d.isnull()['cat_not_in_both'])
def test_bhattacharyya_cat_str_input(self):
var = 'cat_same_dist'
d = sf.bhattacharyya(self.df_1, self.df_2, categorical_variables=var)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(var in d.index)
self.assertEqual(0.0, d[var])
def test_bhattacharyya_wrong_number_of_integration_points(self):
cont = ['cont_same_dist', 'cont_diff_dist']
with self.assertRaises(AssertionError) as ctx:
d = sf.bhattacharyya(self.df_1, self.df_2, continuous_variables=cont, continuous_integration_points=1024)
self.assertEqual('The number of integration points must be 2**n+1 where n is a non-negative integer',
str(ctx.exception))
def test_bhattacharyya_no_columns_passed(self):
self.assertIsNone(sf.bhattacharyya(self.df_1, self.df_2))
def test_bhattacharyya_wrong_type_continuous_variables(self):
with self.assertRaises(AssertionError) as ctx:
sf.bhattacharyya(self.df_1, self.df_2, continuous_variables=123)
self.assertEqual("continuous_variables must be a string, a list or None", str(ctx.exception))
def test_bhattacharyya_wrong_type_df_2(self):
with self.assertRaises(AssertionError) as ctx:
sf.bhattacharyya(self.df_1, self.str_2)
self.assertEqual("df_2 must be a pandas DataFrame", str(ctx.exception))
def test_bhattacharyya_missing_one_input(self):
with self.assertRaises(TypeError):
sf.bhattacharyya()
def test_bhattacharyya_cont_non_numeric_column(self):
var = ['cont_same_dist', 'cont_diff_dist', 'cont_not_in_both']
df = self.df_2.assign(cont_not_in_both=np.random.choice(['a','b','c'], size=self.df_2.shape[0]))
with self.assertRaises(AssertionError) as ctx:
sf.bhattacharyya(self.df_1, df, continuous_variables=var)
self.assertEqual('All continuous variables must be numerical', str(ctx.exception))
def test_bhattacharyya_cont_str_input(self):
var = 'cont_same_dist'
d = sf.bhattacharyya(self.df_1, self.df_2, continuous_variables=var)
self.assertTrue(isinstance(d, pd.Series))
self.assertTrue(var in d.index)
self.assertEqual(0.0138, np.round(d[var], decimals=4))
