def test_stabilize_error(self, df_c):
ipsw = IPSW(df_c, exposure='A', outcome='Y', selection='S')
with pytest.raises(ValueError):
ipsw.sampling_model('L + W_sq',
model_numerator='W',
stabilized=False,
print_results=False)
def test_no_model_error(self, df_c):
ipsw = IPSW(df_c,
exposure='A',
outcome='Y',
selection='S',
generalize=True)
with pytest.raises(ValueError):
ipsw.fit()
def test_transport_stabilized(self, df_r):
ipsw = IPSW(df_r,
exposure='A',
outcome='Y',
selection='S',
generalize=False)
ipsw.sampling_model('L + W_sq', print_results=False)
ipsw.fit()
npt.assert_allclose(ipsw.risk_difference, 0.034896, atol=1e-5)
npt.assert_allclose(ipsw.risk_ratio, 1.097139, atol=1e-4)
def test_transport_iptw(self, df_c):
ipsw = IPSW(df_c,
exposure='A',
outcome='Y',
selection='S',
generalize=False)
ipsw.sampling_model('L + W + W_sq', print_results=False)
ipsw.treatment_model('L', print_results=False)
ipsw.fit()
npt.assert_allclose(ipsw.risk_difference, 0.047296, atol=1e-5)
npt.assert_allclose(ipsw.risk_ratio, 1.1372, atol=1e-4)
def test_generalize_weight(self, df_c):
ipsw = IPSW(df_c,
exposure='A',
outcome='Y',
selection='S',
generalize=True,
weights='weight')
ipsw.sampling_model('L + W + W_sq', print_results=False)
ipsw.treatment_model('L', print_results=False)
ipsw.fit()
npt.assert_allclose(ipsw.risk_difference, 0.055034, atol=1e-5)
npt.assert_allclose(ipsw.risk_ratio, 1.167213, atol=1e-4)
def test_generalize_stabilized(self, df_r):
ipsw = IPSW(df_r, exposure='A', outcome='Y', selection='S')
ipsw.sampling_model('L + W_sq', stabilized=True, print_results=False)
ipsw.fit()
npt.assert_allclose(ipsw.risk_difference, 0.046809, atol=1e-5)
npt.assert_allclose(ipsw.risk_ratio, 1.13905, atol=1e-4)
def test_transport_iptw(self, df_iptw):
ipsw = IPSW(df_iptw, exposure='A', outcome='Y', selection='S', generalize=False, weights='iptw')
ipsw.regression_models('L + W + W_sq', print_results=False)
ipsw.fit()
npt.assert_allclose(ipsw.risk_difference, 0.047296, atol=1e-5)
npt.assert_allclose(ipsw.risk_ratio, 1.1372, atol=1e-4)