def test_group_targets(self):
data = pd.read_csv(self.path_data_B)
meta = json.load(open(self.path_meta_B))
weight = '_'.join(['weights', self.scheme_name_A3])
# Run weights for scheme_A3
engine_B = WeightEngine(data=data, meta=meta)
engine_B.add_scheme(scheme=self.scheme_A3,
key='identity',
verbose=False)
engine_B.run()
data_A3 = engine_B.dataframe("scheme_name_A3")
# check identical weighted column frequencies
df = data_A3.pivot_table(values=[weight],
index=['profile_gender'],
columns=['age_group'],
aggfunc='sum')
for column in df.columns.tolist():
self.assertTrue(
numpy.allclose(df[column].values, numpy.array([1.645, 1.855])))
# check weighted group frequencies have equal proportions
values = data_A3.pivot_table(values=[weight],
index=['age_group'],
aggfunc='sum').values
self.assertTrue(numpy.allclose(values, 3.5))
def test_wdf_structure(self):
data = pd.read_csv(self.path_data_exA)
engine = WeightEngine(data)
targets_gender = [45.6, 54.4]
targets_locality = [10, 15, 20, 25, 30]
weight_targets = [{
'gender':
{code: prop
for code, prop in enumerate(targets_gender, start=1)}
}, {
'locality': {
code: prop
for code, prop in enumerate(targets_locality, start=1)
}
}]
scheme = Rim('complex_filter')
scheme.add_group(name='W1, male',
filter_def='Wave==1 & religion==1',
targets=weight_targets)
scheme.add_group(name='W2, female',
filter_def='Wave==2 & religion==2',
targets=weight_targets)
engine.add_scheme(scheme, key='unique_id', verbose=False)
engine.run()
wdf = engine.dataframe('complex_filter')
self.assertEqual(
sorted(wdf.columns.tolist()),
sorted([
'unique_id', 'gender', 'locality', 'weights_complex_filter',
'Wave', 'religion'
]))
self.assertTrue(len(wdf.index) == 596)