def test_data_conversion(self, flat_data):
ipc = IPCW(flat_data, idvar='id', time='t', event='Y', flat_df=True)
expected_data = pd.DataFrame.from_records([{'id': 1, 't_enter': 0, 't_out': 1, 'A': 1, 'Y': 0.0,
'__uncensored__': 0},
{'id': 2, 't_enter': 0, 't_out': 1, 'A': 1, 'Y': 0.0,
'__uncensored__': 1},
{'id': 2, 't_enter': 1, 't_out': 2, 'A': 1, 'Y': 0.0,
'__uncensored__': 0},
{'id': 3, 't_enter': 0, 't_out': 1, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 3, 't_enter': 1, 't_out': 2, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 3, 't_enter': 2, 't_out': 3, 'A': 0, 'Y': 1.0,
'__uncensored__': 1},
{'id': 4, 't_enter': 0, 't_out': 1, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 4, 't_enter': 1, 't_out': 2, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 4, 't_enter': 2, 't_out': 3, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 5, 't_enter': 0, 't_out': 1, 'A': 0, 'Y': 0.0,
'__uncensored__': 1},
{'id': 5, 't_enter': 1, 't_out': 2, 'A': 0, 'Y': 1.0,
'__uncensored__': 1}]
)
pdt.assert_frame_equal(ipc.df[['id', 'A', 'Y', 't_enter', 't_out', '__uncensored__']],
expected_data[['id', 'A', 'Y', 't_enter', 't_out', '__uncensored__']],
check_dtype=False, check_index_type=False, check_like=True)
def test_data_late_entry_error(self):
df = pd.DataFrame()
df['id'] = [1, 1, 2, 2, 3, 4, 4, 5, 5]
df['t'] = [1, 2, 1, 2, 2, 1, 2, 1, 2]
df['Y'] = [0, 0, 0, 1, 0, 0, 1, 0, 1]
with pytest.raises(ValueError):
IPCW(df, idvar='id', time='t', event='Y', flat_df=False)
def test_data_conversion_late_entry(self, flat_data2):
with pytest.raises(ValueError):
IPCW(flat_data2,
idvar='id',
time='t',
event='Y',
enter='enter',
flat_df=True)
def test_match_sas_weights(self):
sas_w_mean = 0.9993069
sas_w_max = 1.7980410
sas_w_min = 0.8986452
df = load_sample_data(timevary=True)
df['cd40_q'] = df['cd40'] ** 2
df['cd40_c'] = df['cd40'] ** 3
df['cd4_q'] = df['cd4'] ** 2
df['cd4_c'] = df['cd4'] ** 3
df['enter_q'] = df['enter'] ** 2
df['enter_c'] = df['enter'] ** 3
df['age0_q'] = df['age0'] ** 2
df['age0_c'] = df['age0'] ** 3
ipc = IPCW(df, idvar='id', time='enter', event='dead')
cmodeln = 'enter + enter_q + enter_c'
cmodeld = '''enter + enter_q + enter_c + male + age0 + age0_q + age0_c + dvl0 + cd40 +
cd40_q + cd40_c + dvl + cd4 + cd4_q + cd4_c'''
ipc.regression_models(model_denominator=cmodeld, model_numerator=cmodeln)
ipc.fit()
cw = ipc.Weight
npt.assert_allclose(np.mean(cw), sas_w_mean)
npt.assert_allclose(np.max(cw), sas_w_max)
npt.assert_allclose(np.min(cw), sas_w_min)
def test_data_conversion_warning(self, flat_data):
with pytest.warns(UserWarning):
IPCW(flat_data, idvar='id', time='t', event='Y', flat_df=True)
def test_error_time_is_only_one(self, edata2):
with pytest.raises(ValueError):
IPCW(edata2, idvar='id', time='t', event='Y')
def test_error_missing_times(self, edata):
with pytest.raises(ValueError):
IPCW(edata, idvar='id', time='t', event='Y')