def test_by_month(self):
"""
Tests the find_period_ranges function when the data is resampled by month.
"""
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace',
'Pioneer Square', 'Interbay', 'Mann', 'Minor'
]
transition_date = "2016-10-02"
out_df = render_stats.pivot_by_treatment(
DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
resample_by='M',
agg_by='injuries')
out = render_stats.find_period_ranges(out_df,
transition_date=transition_date)
#Test min date
min_date = out_df.index.min()
min_date = datetime.date(min_date.year, min_date.month, min_date.day)
self.assertTrue(min_date.strftime('%Y-%m-%d') == out[0][0])
#Test transition date
rounded_transition_date = "2016-10-31"
self.assertTrue(rounded_transition_date == out[1][0])
#Test max date
max_date = out_df.index.max()
max_date = datetime.date(max_date.year, max_date.month, max_date.day)
self.assertTrue(max_date.strftime('%Y-%m-%d') == out[1][1])
def test_treatment_control(self):
"""
Tests that the correct number of treatment and control values are returned from
pivot_by_treatment given a list of treatment neighborhoods and a list of
control neighborhoods.
"""
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace'
]
test_control_in = ['Pioneer Square', 'Interbay', 'Mann', 'Minor']
out = render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
control_list=test_control_in,
resample_by='D')
change_speed_limit_object_count = int(len(test_treatment_in))
same_speed_limit_object_count = int(len(test_control_in))
self.assertTrue(
int(out.sum()['SpeedLimitChange']) ==
change_speed_limit_object_count)
self.assertTrue(
int(out.sum()['SpeedLimitSame']) == same_speed_limit_object_count)
def test_value_errors(self):
"""
Tests ValueError scenarios in pivot_by_treatment_value.
These include:
1) trying to pivot on a dataframe with no neighborhood object_id
2) passing a bad path for the neighborhood json file
3) attempting to resample the data by year
4) trying to sum by a non-string column
5) trying to sum by a column which doesn't exist
"""
#Test dataframe without neighborhood
with self.assertRaises(ValueError):
render_stats.pivot_by_treatment(DF_NO_NEIGHBORHOODS,
treatment_list=['Genesee'])
#Test bad path
with self.assertRaises(ValueError):
render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=['Genesee'],
neighborhood_path='bad_path')
#Test invalid resample_by
with self.assertRaises(ValueError):
render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=['Genesee'],
resample_by='Y')
#Test agg_by not a string
with self.assertRaises(ValueError):
render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=['Genesee'],
agg_by=12)
#Test agg_by not a column
with self.assertRaises(ValueError):
render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=['Genesee'],
agg_by='fake_column')
def test_agg_by(self):
"""
Tests the pivot_by_treatment functionality to resample the data by month, instead of day.
"""
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace',
'Pioneer Square', 'Interbay', 'Mann', 'Minor'
]
out = render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
resample_by='D',
agg_by='injuries')
injury_count = int(out.sum()['SpeedLimitSame'])
self.assertTrue(injury_count == 21)
def test_resample_month(self):
"""
Tests the pivot_by_treatment functionality to resample the data by month, instead of day.
"""
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace',
'Pioneer Square', 'Interbay', 'Mann', 'Minor'
]
out = render_stats.pivot_by_treatment(DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
resample_by='M')
self.assertTrue(out.index.min().month == 3)
self.assertTrue(out.index.min().day == 31)
def test_transition_date_before_start(self):
"""
Tests that a ValueError is returned when transition_date before falls before any date in the
data.
"""
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace',
'Pioneer Square', 'Interbay', 'Mann', 'Minor'
]
transition_date = "1900-01-01"
out_df = render_stats.pivot_by_treatment(
DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
resample_by='D',
agg_by='injuries')
with self.assertRaises(ValueError):
render_stats.find_period_ranges(out_df,
transition_date=transition_date)
def test_causal_impact_load(self):
"""
Performs a smoke test where we load causal impact.
"""
#Setup
test_treatment_in = [
'Atlantic', 'Pike-Market', 'Belltown', 'International District',
'Central Business District', 'First Hill', 'Yesler Terrace',
'Pioneer Square', 'Interbay', 'Mann', 'Minor'
]
transition_date = "2016-12-22"
out_df = render_stats.pivot_by_treatment(
DF_NEIGHBORHOODS,
treatment_list=test_treatment_in,
resample_by='D',
agg_by=None)
out_date = render_stats.find_period_ranges(
out_df, transition_date=transition_date)
#Call causal impact package
causal_impact_out = CausalImpact(out_df, out_date[0], out_date[1])
self.assertTrue(isinstance(causal_impact_out, CausalImpact))