def testDuplicateGeoInSamePair(self):
temp_df = self.df.copy()
# change geo #1 to geo #2, so that geo #2 is duplicated in pair 1 with
# assignment 0 and 1
temp_df.loc[temp_df['geo'] == 1, 'geo'] = 2
with self.assertRaisesRegex(
ValueError, r'Some geos are duplicated and appear in multiple pairs.'):
trimmed_match_post_analysis.check_input_data(temp_df)
def testCheckGeosPerPairInInputData(self):
temp_df = self.df.copy()
# reassign geo #2 which is in treatment, and geo #3 which is control
temp_df.loc[temp_df['geo'] == 2, 'assignment'] = 0
temp_df.loc[temp_df['geo'] == 3, 'assignment'] = 1
with self.assertRaisesRegex(
ValueError, r'Some pairs do not have one geo for each group.'):
trimmed_match_post_analysis.check_input_data(temp_df)
def testDuplicateGeoInInputData(self):
temp_df = self.df.copy()
# change geo #4 to geo #2, so that geo #2 is duplicated
temp_df.loc[temp_df['geo'] == 4, 'geo'] = 2
# change geo #5 to geo #3, so that geo #3 is duplicated
temp_df.loc[temp_df['geo'] == 5, 'geo'] = 3
with self.assertRaisesRegex(
ValueError, r'Some geos are duplicated and appear in multiple pairs.'):
trimmed_match_post_analysis.check_input_data(temp_df)
def testCheckInputDataColumns(self):
temp_df = self.df.copy()
# remove the column assignment
temp_df.drop(columns='assignment', inplace=True)
with self.assertRaisesRegex(
ValueError,
'The mandatory columns {\'assignment\'} are missing from the input data'
):
trimmed_match_post_analysis.check_input_data(temp_df)
def testCheckInputDataCorrectGroupLabels(self):
with self.assertRaisesRegex(
ValueError,
r'The data do not have observations for the two groups.' +
r'Check the data and the values used to indicate the ' +
r'assignments for treatment and control. The labels ' +
r'found in the data in input are ' +
r'\[0, 1\], and the expected labels ' +
r'are: Treatment=1, Control=2'):
trimmed_match_post_analysis.check_input_data(
self.df, group_control=2, group_treatment=1)
def testManyGeosPerPairWithSameAssignmentInInputData(self):
temp_df = self.df.copy()
# add one additional geo to pairs #1 and #4
temp_df = temp_df.append(pd.DataFrame({
'date': ['2020-10-10', '2020-10-10'],
'geo': [10, 9],
'response': [10, 11],
'cost': [1.0, 2.0],
'pair': [1, 4],
'assignment': [0, 1],
'period': [1, 1],
}))
with self.assertRaisesRegex(
ValueError, r'Some pairs do not have one geo for each group.'):
trimmed_match_post_analysis.check_input_data(temp_df)
def testCheckUnequalGroupSizesInInputData(self):
temp_df = self.df.copy()
# remove geo #2 which is in treatment
temp_df = temp_df[temp_df['geo'] != 2]
with self.assertRaisesRegex(ValueError,
r'Some pairs do not have one geo for ' +
r'each group.'):
trimmed_match_post_analysis.check_input_data(temp_df)
temp_df = self.df.copy()
# remove geo #1 which is in control
temp_df = temp_df[temp_df['geo'] != 1]
with self.assertRaisesRegex(ValueError,
r'Some pairs do not have one geo for ' +
r'each group.'):
trimmed_match_post_analysis.check_input_data(temp_df)
def testCheckInputData(self):
temp_df = self.df.copy()
# remove one observation for geo #2
temp_df = temp_df[~((temp_df['geo'] == 2) &
(temp_df['date'] == '2020-10-10'))]
geox_data = trimmed_match_post_analysis.check_input_data(temp_df)
expected_df = pd.DataFrame({
'date': ['2020-10-09', '2020-10-10', '2020-10-11'] * 4,
'geo': [1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4],
'pair': [1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2],
'assignment': [0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1],
'response': [10, 10, 10, 20, 0.0, 20, 30, 30, 30, 40, 40, 40],
'cost': [1.0, 1.0, 1.0, 2.0, 0.0, 2.0, 3.0, 3.0, 3.0, 4.0, 4.0, 4.0],
}).sort_values(by=['date', 'geo']).reset_index(drop=True)
self.assertTrue(geox_data.equals(expected_df))
def testCheckInputDataOrder(self):
temp = pd.DataFrame({
'date': pd.date_range('2021-01-01', periods=4),
'pair': [2, 2, 1, 1],
'assignment': [1, 2, 2, 1],
'geo': [1, 2, 3, 4],
'response': 0,
'cost': 0
})
geox_data = trimmed_match_post_analysis.check_input_data(
temp, group_control=1, group_treatment=2)
self.assertTrue(
np.array_equal(
geox_data[['date', 'geo', 'pair', 'assignment']],
pd.DataFrame({
'date': np.repeat(temp['date'], 4),
'geo': [1, 2, 3, 4] * 4,
'pair': [2, 2, 1, 1] * 4,
'assignment': [1, 2, 2, 1] * 4
})))
def testGeosNotInExperimentAreExcluded(self):
temp_df = self.df.copy()
# add two additional geos with assignment -1 in new and different pairs.
temp_df = temp_df.append(pd.DataFrame({
'date': ['2020-10-10', '2020-10-10'],
'geo': [9, 10],
'response': [10, 11],
'cost': [1.0, 2.0],
'pair': [100, 101],
'assignment': [-1, -1],
}))
geox_data = trimmed_match_post_analysis.check_input_data(temp_df)
expected_df = pd.DataFrame({
'date': ['2020-10-09', '2020-10-10', '2020-10-11'] * 4,
'geo': [1, 1, 1, 2, 2, 2, 3, 3, 3, 4, 4, 4],
'pair': [1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2],
'assignment': [0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1],
'response': [10, 10, 10, 20, 20.0, 20, 30, 30, 30, 40, 40, 40],
'cost': [1.0, 1.0, 1.0, 2.0, 2.0, 2.0, 3.0, 3.0, 3.0, 4.0, 4.0, 4.0],
}).sort_values(by=['date', 'geo']).reset_index(drop=True)
self.assertTrue(geox_data.reset_index(drop=True).equals(expected_df))
